<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom" xmlns:dc="http://purl.org/dc/elements/1.1/">
  <channel>
    <title>DEV Community: Skapi</title>
    <description>The latest articles on DEV Community by Skapi (@skapi_api).</description>
    <link>https://dev.to/skapi_api</link>
    <image>
      <url>https://media2.dev.to/dynamic/image/width=90,height=90,fit=cover,gravity=auto,format=auto/https:%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Fuser%2Fprofile_image%2F3161077%2Ff42727c1-df28-406e-85bd-b25b4a006818.png</url>
      <title>DEV Community: Skapi</title>
      <link>https://dev.to/skapi_api</link>
    </image>
    <atom:link rel="self" type="application/rss+xml" href="https://dev.to/feed/skapi_api"/>
    <language>en</language>
    <item>
      <title>DDBD: free and open source database on Linux and macOS</title>
      <dc:creator>Skapi</dc:creator>
      <pubDate>Fri, 13 Nov 2026 10:06:32 +0000</pubDate>
      <link>https://dev.to/skapi_api/ddbd-free-and-open-source-database-on-linux-and-macos-9e0</link>
      <guid>https://dev.to/skapi_api/ddbd-free-and-open-source-database-on-linux-and-macos-9e0</guid>
      <description>&lt;h2&gt;
  
  
  DDBD: the database behind Skapi, now free and open source on Linux and macOS
&lt;/h2&gt;

&lt;p&gt;Hello, and thank you for being here. Today we are releasing something we have been building toward since the first day of Skapi: &lt;strong&gt;DDBD&lt;/strong&gt;, a schemaless record database with writer-based access control (WBAC). It is the record design that runs Skapi's database, packed into a server you can run yourself. It is free and open source under the Apache-2.0 license, and it runs on Linux and macOS. Everything about it is at &lt;a href="https://ddbd.skapi.com" rel="noopener noreferrer"&gt;ddbd.skapi.com&lt;/a&gt;.&lt;/p&gt;

&lt;p&gt;This article explains the problems we set out to solve from day one, how DDBD solves them, how it performs against PostgreSQL, MongoDB and SQLite, and how to install it in a minute.&lt;/p&gt;

&lt;h2&gt;
  
  
  The problems we set out to solve from day one
&lt;/h2&gt;

&lt;p&gt;When we started Skapi, we wanted four things from a database, all at the same time.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Records without a schema.&lt;/strong&gt; You should be able to save a JSON record and query it right away, without designing tables, writing migrations or setting up indexes first. In DDBD, the user who saves a record names its collection, its access level, its tags and its indexes in the same request. There is nothing to set up beforehand.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Access rules that belong to the data.&lt;/strong&gt; Most databases ask an administrator to write and maintain access policies for every user and every case. We wanted the opposite: the person who writes a record decides who can read it, and the database enforces it on every read. We call this &lt;strong&gt;writer-based access control (WBAC)&lt;/strong&gt;. RBAC asks who you are. WBAC asks who wrote it. A record can be private to its writer, shared with users at an access level, or readable by specific users, and no admin has to keep a policy up to date.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Fully serverless, scaling at low cost.&lt;/strong&gt; We wanted a backend with no database server to keep running around the clock, where the cost of a request follows the size of its answer, not the size of the database. That only works if every query is cheap by construction, with no query planner deciding at read time how to combine indexes.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Safer AI agent development.&lt;/strong&gt; Agents now read and write data on their own. An agent should get only what its user may see, within limits the user sets, and everything it writes should say which agent wrote it. Nobody should have to hand an agent a password, or an admin key.&lt;/p&gt;

&lt;h2&gt;
  
  
  The core technology that runs Skapi
&lt;/h2&gt;

&lt;p&gt;The answer to all four is one design. When a record is saved, its writer's choices (collection, access level, tag, index value) are turned into sorted keys and written together with the record, in one atomic step. A query then reads one range of those keys for each access level the reader may read, merged in order. It starts at the first match and stops when the page is full.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fo4be6y0g1q49uqqegqt1.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fo4be6y0g1q49uqqegqt1.png" alt="access rule" width="800" height="380"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;&lt;em&gt;The access rule is part of the key. A reader's query never reaches records it may not read, so there is nothing to filter out after the fact.&lt;/em&gt;&lt;/p&gt;

&lt;p&gt;This is what lets Skapi run with no database server to keep running: every &lt;code&gt;getRecords()&lt;/code&gt; call on Skapi works this way, as one range read on a sorted key store, so it runs in short-lived functions that cost nothing while idle and scale with the traffic. Access control costs nothing extra at read time, because it is already in the key. And schemaless records cost nothing extra either, because the writer, not a schema, says what to index.&lt;/p&gt;

&lt;p&gt;DDBD is this same record design as a standalone database server, on RocksDB. It speaks HTTP and JSON, so any language with an HTTP client can use it without a client library, and it comes with a user shell and a web admin.&lt;/p&gt;

&lt;h2&gt;
  
  
  How fast it is
&lt;/h2&gt;

&lt;p&gt;We measured DDBD 0.1.2 against PostgreSQL 16.2, MongoDB 8.0.4 and SQLite 3.46.0 with the same million products in each. Every product has a price, one of 20 tags (&lt;code&gt;sale&lt;/code&gt; on 5% of them) and an access level. The query asks for the 50 cheapest products tagged &lt;code&gt;sale&lt;/code&gt; above a random price, the kind of question an app or an agent asks all day, for a user who may read only part of the records.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fhrwniyydbf216ua1h74u.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fhrwniyydbf216ua1h74u.png" alt="PostgreSQL" width="800" height="333"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;&lt;em&gt;With 1,000,000 records and a user who may read 1% of them, DDBD answered in 2.7 ms: 5.8× faster than PostgreSQL, 6.7× faster than MongoDB and 24× faster than SQLite.&lt;/em&gt;&lt;/p&gt;

&lt;p&gt;When the user may read only 0.1% of the records, the gap is similar: DDBD 1.8 ms, PostgreSQL 10.8 ms, MongoDB 4.0 ms, SQLite 65.8 ms.&lt;/p&gt;

&lt;p&gt;The more important result is what happens as the database grows. From 10,000 to 1,000,000 records, a hundred times more data, the same query took 6.8× longer on DDBD, while it took 24× longer on MongoDB, 30× longer on PostgreSQL and 98× longer on SQLite. Part of DDBD's growth is the answer itself: at 10,000 records only 4.1 products matched on average, at 1,000,000 the page was full with 46.3.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fdf00zij9z1yw89qd8uxk.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fdf00zij9z1yw89qd8uxk.png" alt="Query time" width="800" height="455"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;&lt;em&gt;Query time grows with the answer, not with the database.&lt;/em&gt;&lt;/p&gt;

&lt;h3&gt;
  
  
  All results at 1,000,000 records
&lt;/h3&gt;

&lt;div class="table-wrapper-paragraph"&gt;&lt;table&gt;
&lt;thead&gt;
&lt;tr&gt;
&lt;th&gt;Query&lt;/th&gt;
&lt;th&gt;Caller may read&lt;/th&gt;
&lt;th&gt;Records returned&lt;/th&gt;
&lt;th&gt;DDBD&lt;/th&gt;
&lt;th&gt;PostgreSQL&lt;/th&gt;
&lt;th&gt;MongoDB&lt;/th&gt;
&lt;th&gt;SQLite&lt;/th&gt;
&lt;/tr&gt;
&lt;/thead&gt;
&lt;tbody&gt;
&lt;tr&gt;
&lt;td&gt;tag &lt;code&gt;sale&lt;/code&gt; and price ≥ X&lt;/td&gt;
&lt;td&gt;0.1%&lt;/td&gt;
&lt;td&gt;21.7&lt;/td&gt;
&lt;td&gt;&lt;strong&gt;1.8 ms&lt;/strong&gt;&lt;/td&gt;
&lt;td&gt;10.8 ms&lt;/td&gt;
&lt;td&gt;4.0 ms&lt;/td&gt;
&lt;td&gt;65.8 ms&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;tag &lt;code&gt;sale&lt;/code&gt; and price ≥ X&lt;/td&gt;
&lt;td&gt;1%&lt;/td&gt;
&lt;td&gt;46.3&lt;/td&gt;
&lt;td&gt;&lt;strong&gt;2.7 ms&lt;/strong&gt;&lt;/td&gt;
&lt;td&gt;15.8 ms&lt;/td&gt;
&lt;td&gt;18.2 ms&lt;/td&gt;
&lt;td&gt;64.7 ms&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;tag &lt;code&gt;sale&lt;/code&gt; and price ≥ X&lt;/td&gt;
&lt;td&gt;10%&lt;/td&gt;
&lt;td&gt;50&lt;/td&gt;
&lt;td&gt;&lt;strong&gt;3.5 ms&lt;/strong&gt;&lt;/td&gt;
&lt;td&gt;12.1 ms&lt;/td&gt;
&lt;td&gt;30.9 ms&lt;/td&gt;
&lt;td&gt;68.2 ms&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;tag &lt;code&gt;sale&lt;/code&gt; and price ≥ X&lt;/td&gt;
&lt;td&gt;100%&lt;/td&gt;
&lt;td&gt;50&lt;/td&gt;
&lt;td&gt;4.9 ms&lt;/td&gt;
&lt;td&gt;&lt;strong&gt;3.1 ms&lt;/strong&gt;&lt;/td&gt;
&lt;td&gt;6.0 ms&lt;/td&gt;
&lt;td&gt;65.2 ms&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;price ≥ X, no tag&lt;/td&gt;
&lt;td&gt;0.1%&lt;/td&gt;
&lt;td&gt;48.1&lt;/td&gt;
&lt;td&gt;&lt;strong&gt;2.1 ms&lt;/strong&gt;&lt;/td&gt;
&lt;td&gt;4.5 ms&lt;/td&gt;
&lt;td&gt;4.1 ms&lt;/td&gt;
&lt;td&gt;92.3 ms&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;price ≥ X, no tag&lt;/td&gt;
&lt;td&gt;1%&lt;/td&gt;
&lt;td&gt;49.9&lt;/td&gt;
&lt;td&gt;&lt;strong&gt;3.1 ms&lt;/strong&gt;&lt;/td&gt;
&lt;td&gt;7.5 ms&lt;/td&gt;
&lt;td&gt;16.2 ms&lt;/td&gt;
&lt;td&gt;10.6 ms&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;price ≥ X, no tag&lt;/td&gt;
&lt;td&gt;10%&lt;/td&gt;
&lt;td&gt;50&lt;/td&gt;
&lt;td&gt;3.6 ms&lt;/td&gt;
&lt;td&gt;&lt;strong&gt;2.5 ms&lt;/strong&gt;&lt;/td&gt;
&lt;td&gt;4.3 ms&lt;/td&gt;
&lt;td&gt;&lt;strong&gt;2.5 ms&lt;/strong&gt;&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;price ≥ X, no tag&lt;/td&gt;
&lt;td&gt;100%&lt;/td&gt;
&lt;td&gt;50&lt;/td&gt;
&lt;td&gt;3.7 ms&lt;/td&gt;
&lt;td&gt;2.0 ms&lt;/td&gt;
&lt;td&gt;2.9 ms&lt;/td&gt;
&lt;td&gt;&lt;strong&gt;1.6 ms&lt;/strong&gt;&lt;/td&gt;
&lt;/tr&gt;
&lt;/tbody&gt;
&lt;/table&gt;&lt;/div&gt;

&lt;p&gt;&lt;em&gt;The fastest time in each row is in bold.&lt;/em&gt;&lt;/p&gt;

&lt;p&gt;The table shows where each design does its work. With a tag and an access level in the query, a SQL database or MongoDB has to combine the tag index, the price order and the access filter at query time, and the less a user may read, the more it has to skip. DDBD wrote a key that already holds all three when the record was saved. A plain price range with no tag, for a user who may read 10% or more of the records, is exactly what a single price index serves best, and there PostgreSQL and SQLite answer faster. When the user may read every record there is no access filter to combine, and PostgreSQL answers the tag query faster too. That first kind is the everyday work of an app with many users, or of agents that each read their own slice of a shared store.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;How it was measured.&lt;/strong&gt; An Intel Core i7-1160G7 laptop (4 cores, 8 threads, 15 GB RAM) with all data in memory. Every query was the same HTTP &lt;code&gt;POST&lt;/code&gt; with a JSON body and a session token to each database, and every answer was the 50 full records as JSON. PostgreSQL, MongoDB and SQLite answered through a small HTTP server written for the test with the same libraries as DDBD, which reads the token, looks up the user's access level, runs the query and writes the answer. Each database had the indexes that serve the query, 300 queries per case after 30 warm-up queries, the four taking turns query by query, and the whole set ran 3 times; each number is the median of the runs' medians. All four returned the same records in the same order for every query.&lt;/p&gt;

&lt;h2&gt;
  
  
  Built for AI agents
&lt;/h2&gt;

&lt;p&gt;A DDBD user makes an &lt;strong&gt;agent key&lt;/strong&gt; for each agent or script, without asking an admin, and can revoke it at any time. A key can be limited to some collections, to a highest access level, to read-only, to private records or not, and to an expiry date:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight json"&gt;&lt;code&gt;&lt;span class="err"&gt;//&lt;/span&gt; &lt;span class="n"&gt;createAgentKey&lt;/span&gt; &lt;span class="n"&gt;body&lt;/span&gt;
&lt;span class="p"&gt;{&lt;/span&gt;
  &lt;span class="nl"&gt;"name"&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="s2"&gt;"shop-bot"&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt;
  &lt;span class="nl"&gt;"collections"&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="p"&gt;[&lt;/span&gt;&lt;span class="s2"&gt;"products"&lt;/span&gt;&lt;span class="p"&gt;]&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt;
  &lt;span class="nl"&gt;"max_level"&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="s2"&gt;"05"&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt;
  &lt;span class="nl"&gt;"read_only"&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="kc"&gt;true&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt;
  &lt;span class="nl"&gt;"expires"&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="s2"&gt;"30d"&lt;/span&gt;
&lt;span class="p"&gt;}&lt;/span&gt;
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;An agent reads only what its user may read, within the key's limits. Every record it saves belongs to the user and names the key that saved it:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight json"&gt;&lt;code&gt;&lt;span class="p"&gt;{&lt;/span&gt;
  &lt;span class="nl"&gt;"record_id"&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="s2"&gt;"VZYu683bLzapjrgk"&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt;
  &lt;span class="nl"&gt;"user_id"&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="s2"&gt;"JFeCedoO"&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt;
  &lt;span class="nl"&gt;"agent"&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="p"&gt;{&lt;/span&gt; &lt;span class="nl"&gt;"key_id"&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="s2"&gt;"mtxBMs"&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="nl"&gt;"name"&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="s2"&gt;"shop-bot"&lt;/span&gt; &lt;span class="p"&gt;}&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt;
  &lt;span class="nl"&gt;"data"&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="p"&gt;{&lt;/span&gt; &lt;span class="nl"&gt;"title"&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="s2"&gt;"Desk Lamp"&lt;/span&gt; &lt;span class="p"&gt;}&lt;/span&gt;
&lt;span class="p"&gt;}&lt;/span&gt;
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;Two more things make it safe to let agents work on shared data:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;No network user is ever an admin.&lt;/strong&gt; The only administrator is the operating system user who runs the server, through a local socket. A leaked token, at any level, never carries admin powers.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Data even the admin cannot read.&lt;/strong&gt; DDBD gives every user a salt and the settings to derive a key from their password with Argon2id. The app encrypts the data on the user's device before saving it, so the server, its backups and its admin only ever hold the encrypted result.&lt;/li&gt;
&lt;/ul&gt;

&lt;h2&gt;
  
  
  Try it on your machine
&lt;/h2&gt;

&lt;p&gt;DDBD runs on Linux and macOS. On Linux, install it with one command:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight shell"&gt;&lt;code&gt;curl -fsSL https://cdn.broadwayinc.com/get/ddbd/v1 &lt;span class="o"&gt;|&lt;/span&gt; sh
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;On macOS, install it with &lt;a href="https://brew.sh" rel="noopener noreferrer"&gt;Homebrew&lt;/a&gt;:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight shell"&gt;&lt;code&gt;brew install broadwayinc/ddbd/ddbd
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;Then create a data folder and start the server:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight shell"&gt;&lt;code&gt;ddbd init --data ~/ddbd-data
ddbd start --data ~/ddbd-data
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;In a second terminal, turn on the web admin by giving it a password, then open &lt;code&gt;http://localhost:7070/&lt;/code&gt; in your browser. From there you can add your first user, save records, run queries and make agent keys:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight shell"&gt;&lt;code&gt;ddbd admin --data ~/ddbd-data web-password
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;The server speaks plain HTTP by default, which is fine on your own machine. Anywhere else, run it behind HTTPS, or start it with &lt;code&gt;--tls&lt;/code&gt;. Running the install command again later updates DDBD to the newest version; with Homebrew, &lt;code&gt;brew upgrade ddbd&lt;/code&gt; does.&lt;/p&gt;

&lt;p&gt;The documentation covers every request, with curl, JavaScript and Python examples: &lt;a href="https://ddbd.skapi.com" rel="noopener noreferrer"&gt;ddbd.skapi.com&lt;/a&gt;.&lt;/p&gt;

&lt;h2&gt;
  
  
  Thank you
&lt;/h2&gt;

&lt;p&gt;Skapi has always been about letting you build without running a backend yourself, and that does not change: your Skapi projects keep running on the same design, managed by us. DDBD is for when you want that design on your own machine, next to your own agents and your own data, free.&lt;/p&gt;

&lt;p&gt;Thank you for building with Skapi, and for every question, bug report and idea you send us. Try DDBD, break it, and tell us what you think at &lt;a href="mailto:support@broadwayinc.com"&gt;support@broadwayinc.com&lt;/a&gt;.&lt;/p&gt;

</description>
      <category>database</category>
      <category>opensource</category>
      <category>backend</category>
      <category>backenddevelopment</category>
    </item>
    <item>
      <title>How to Cache Your Website Without Breaking Updates (Using Skapi)</title>
      <dc:creator>Skapi</dc:creator>
      <pubDate>Wed, 11 Nov 2026 09:34:14 +0000</pubDate>
      <link>https://dev.to/skapi_api/how-to-cache-your-website-without-breaking-updates-using-skapi-2lkf</link>
      <guid>https://dev.to/skapi_api/how-to-cache-your-website-without-breaking-updates-using-skapi-2lkf</guid>
      <description>&lt;h2&gt;
  
  
  The CDN cache and the browser cache are not the same thing
&lt;/h2&gt;

&lt;p&gt;The question comes in several shapes. "I uploaded the new version and the site still shows the old one." "I changed the setting to one year and now I cannot get my CSS to update." "What is the difference between the CDN cache row and the Browser caching row, and which one do I click?" They are all the same confusion, and it is an understandable one: between the file you upload and the screen your visitor looks at there are two caches, not one, and they belong to two different parties. This article takes them apart, shows how Skapi web hosting uses each, and explains why the way it is set up is the one you would end up choosing yourself after a bad week.&lt;/p&gt;

&lt;h2&gt;
  
  
  Two caches, two owners
&lt;/h2&gt;

&lt;p&gt;A cache is a copy kept close to where it is needed so the original is not fetched again. Web hosting has two of them in a row.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fnara79c300pjlnrmhdmj.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fnara79c300pjlnrmhdmj.png" alt="A cache is a copy" width="800" height="288"&gt;&lt;/a&gt;&lt;br&gt;
&lt;em&gt;Two copies between your upload and your visitor. Skapi owns the first; the visitor's browser owns the second.&lt;/em&gt;&lt;/p&gt;

&lt;p&gt;The &lt;strong&gt;CDN cache&lt;/strong&gt; is Skapi's. The CDN is a network of servers around the world, and when a visitor in Paris asks for your logo, the copy comes from a server near Paris, not from the storage where you uploaded it. That is what makes a hosted site fast everywhere. The copy is kept for up to a day, and because the copy is on our side, we can throw it away whenever your files change. We do, and the next section is about that.&lt;/p&gt;

&lt;p&gt;The &lt;strong&gt;browser cache&lt;/strong&gt; is the visitor's. Once their browser has downloaded a file, it may keep its own copy on the device and reuse it without asking anyone, for as long as the file's headers allow. That copy is on a computer we have never seen. Nothing we do, and nothing you do on the Web Hosting page, reaches it. Only two things ever replace it: the allowed time running out, or the page asking for a different address.&lt;/p&gt;

&lt;p&gt;Hold on to that difference, because it answers most questions by itself. The CDN cache is about how fast the site is for everyone; it is ours to keep fresh, and we do. The browser cache is about how fast a return visit is for one person; it is yours to allow, and once allowed it is out of everyone's hands.&lt;/p&gt;
&lt;h2&gt;
  
  
  The CDN cache: fast, and refreshed for you
&lt;/h2&gt;

&lt;p&gt;Here is what happens when you change something on your site, whether by uploading, overwriting or deleting on the &lt;strong&gt;Web Hosting&lt;/strong&gt; page or through the &lt;a href="https://docs.skapi.com/introduction/getting-started.html#connecting-the-skapi-mcp-server" rel="noopener noreferrer"&gt;Skapi MCP server&lt;/a&gt;.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fdknt3i236ndqf219lk1i.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fdknt3i236ndqf219lk1i.png" alt="Skapi MCP server" width="799" height="267"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;&lt;em&gt;A change refreshes the CDN by itself. The CDN cache row on the Web Hosting page shows where it stands.&lt;/em&gt;&lt;/p&gt;

&lt;p&gt;About a minute after the change finishes, the CDN drops its copies of the whole site. The next visit to any address fetches the new file, and from then on the copies are the new ones. Changes that land during that minute ride the same refresh, so a batch of twenty files is one refresh, not twenty. While it runs, the &lt;strong&gt;CDN cache&lt;/strong&gt; row on the Web Hosting page reads &lt;strong&gt;Updating&lt;/strong&gt;; when it is done it reads &lt;strong&gt;Live&lt;/strong&gt;. There is nothing to click, nothing to configure and nothing to remember.&lt;/p&gt;

&lt;p&gt;Why the whole site and not just the changed file? Because a page names its files. When &lt;code&gt;index.html&lt;/code&gt; changes to point at a new stylesheet and the stylesheet changes too, you want both to arrive together, never a new page with the old stylesheet or the other way round. One refresh per burst of changes keeps the site consistent, and it is also the simplest thing that can possibly work.&lt;/p&gt;

&lt;p&gt;Two safety nets sit behind the automatic refresh. A check runs every 30 minutes and refreshes any site whose files changed without a refresh having happened, so a missed event cannot leave a site stale. And &lt;strong&gt;[Refresh now]&lt;/strong&gt;, on the same row, republishes every file of the site to the CDN at once, for the moment you would rather not wait for anything. One more detail that saves people a support ticket: an address that has no file is never cached, by the CDN or by a browser. Upload a file at a new address and it shows at once, and so does a newly set 404 page.&lt;/p&gt;

&lt;p&gt;A CDN refresh changes what the CDN serves. It does not change what a visitor's browser already keeps. That is the other cache, and the other row.&lt;/p&gt;
&lt;h2&gt;
  
  
  The browser cache: yours to set, per upload
&lt;/h2&gt;

&lt;p&gt;The &lt;strong&gt;Browser caching&lt;/strong&gt; row on the Web Hosting page decides how long a visitor's browser may keep the files you upload before asking for them again.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto/https%3A%2F%2Fcdn.broadwayinc.com%2Farticles%2Fimg%2Fhosting-cache-card.webp" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto/https%3A%2F%2Fcdn.broadwayinc.com%2Farticles%2Fimg%2Fhosting-cache-card.webp" alt="The Web Hosting card: the Address, Storage in Use and 404 Page rows, then the Browser caching row with a choice list reading No caching, and the CDN cache row reading Live with a Refresh now link" width="800" height="421"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;&lt;em&gt;The two rows. Browser caching is a setting you choose; the CDN cache is a state you watch.&lt;/em&gt;&lt;/p&gt;

&lt;p&gt;The choices are &lt;strong&gt;No caching&lt;/strong&gt;, &lt;strong&gt;5 minutes&lt;/strong&gt;, &lt;strong&gt;1 hour&lt;/strong&gt;, &lt;strong&gt;1 day&lt;/strong&gt;, &lt;strong&gt;7 days&lt;/strong&gt;, &lt;strong&gt;30 days&lt;/strong&gt; and &lt;strong&gt;1 year&lt;/strong&gt;, plus &lt;strong&gt;Ask on upload&lt;/strong&gt;, which asks you once per upload and offers to make your answer the default.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;No caching&lt;/strong&gt; is the default, and it is less expensive than it sounds. The browser keeps the file but asks the CDN before every use whether it changed. When it has not, the answer is a tiny "unchanged" and the browser uses its copy; when it has, the new file comes down. The CDN answers from a server near the visitor, so the question is fast, and the site is always current. For a site you are still building, or for any file you edit by hand under the same name, this is the right setting and you never have to think about it.&lt;/p&gt;

&lt;p&gt;A longer setting makes return visits faster, because the browser does not even ask. The right files for it are the ones that never change under their name: images and fonts, a library pinned by version, and the output of a build tool that puts a hash in the file name (&lt;code&gt;app.3f9c1e.js&lt;/code&gt;), which is a new name every time the content changes. For those, &lt;strong&gt;1 year&lt;/strong&gt; is correct and costs nothing. &lt;strong&gt;5 minutes&lt;/strong&gt;, &lt;strong&gt;1 hour&lt;/strong&gt; and &lt;strong&gt;1 day&lt;/strong&gt; are for files that change now and then, where a short delay before visitors see the change is acceptable. A practical way to run a site: keep &lt;strong&gt;No caching&lt;/strong&gt; while it is taking shape, and once it has settled, upload the files that carry a hash or a version in their name with &lt;strong&gt;1 year&lt;/strong&gt;, and leave everything else alone.&lt;/p&gt;

&lt;p&gt;The choice is written onto each file when it is uploaded, and the file list shows it in a column. Files you uploaded before you changed the setting keep what they had; upload a file again to give it the new value.&lt;/p&gt;
&lt;h2&gt;
  
  
  The warning that comes with it
&lt;/h2&gt;

&lt;p&gt;Anything above 5 minutes shows a warning, and &lt;strong&gt;7 days&lt;/strong&gt; or more opens an alert you have to accept. It is there for a reason, and the reason is the sentence from the beginning of this article: the browser cache is the visitor's.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto/https%3A%2F%2Fcdn.broadwayinc.com%2Farticles%2Fimg%2Fhosting-browser-caching-alert.webp" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto/https%3A%2F%2Fcdn.broadwayinc.com%2Farticles%2Fimg%2Fhosting-browser-caching-alert.webp" alt="The Browser caching alert for 7 days: visitors who already have a file will not see a new version of it for 7 days, even after you upload a change; once a visitor's browser has downloaded a file you have no control over it; with Cancel and I understand buttons" width="496" height="478"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;&lt;em&gt;The alert from 7 days up. Read it once, believe it forever.&lt;/em&gt;&lt;/p&gt;

&lt;p&gt;A long browser cache is a one-way door. Once a visitor's browser has downloaded a file with a one-year allowance, that browser will not ask about it again for a year. Uploading a new version does nothing for that visitor. Refreshing the CDN does nothing for that visitor. Deleting the file does nothing for that visitor. You cannot change this for a file with the same name, because the browser will not ask, and a browser that does not ask cannot be told.&lt;/p&gt;

&lt;p&gt;The only way to ship a change to such a file is a new address. Either rename the file, which is what build tools do with the hash, or add a version to the address you use in your pages:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight html"&gt;&lt;code&gt;&lt;span class="nt"&gt;&amp;lt;script&lt;/span&gt; &lt;span class="na"&gt;src&lt;/span&gt;&lt;span class="o"&gt;=&lt;/span&gt;&lt;span class="s"&gt;"/app.js?v=2"&lt;/span&gt;&lt;span class="nt"&gt;&amp;gt;&amp;lt;/script&amp;gt;&lt;/span&gt;
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;&lt;code&gt;?v=&lt;/code&gt; is the one query string the CDN and the browser both treat as a new address. Change the number whenever you change the file, and every visitor gets the new one on their next visit, however long the old one was allowed to live. Any other query string is ignored, on purpose, so that tracking parameters and the like do not fill the caches with duplicates.&lt;/p&gt;

&lt;p&gt;So the rule for a long cache is short: use it only for files whose name or version changes with their content. If you cannot promise that, choose &lt;strong&gt;No caching&lt;/strong&gt; and let the browser ask.&lt;/p&gt;

&lt;h2&gt;
  
  
  Pages are never cached, and that is what makes the rest work
&lt;/h2&gt;

&lt;p&gt;There is one kind of file the setting does not apply to. Every HTML file is served as &lt;strong&gt;No caching&lt;/strong&gt; whatever you choose, and so is the 404 page.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fd29kw4gspugu8j97dxu9.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fd29kw4gspugu8j97dxu9.png" alt="No caching" width="799" height="376"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;&lt;em&gt;Pages always come fresh; the other files keep the choice made at their upload.&lt;/em&gt;&lt;/p&gt;

&lt;p&gt;This is not a limitation. It is the keystone. A page is what names your other files: it says which stylesheet, which script, which images. Because a page is never kept by a browser, a reload always shows the newest page, and the newest page always points at the addresses you put in it. Change &lt;code&gt;app.js&lt;/code&gt; to &lt;code&gt;app.js?v=3&lt;/code&gt; in the page, and every visitor's next reload fetches the new script, even if their browser was holding the old one for a year. The long cache on the files and the no-cache on the pages are two halves of one design. Hashed file names, versioned addresses, the "cache forever" habit of every modern build tool: all of it depends on the page being current, and on Skapi it always is.&lt;/p&gt;

&lt;h2&gt;
  
  
  Why we think this is the right way to do it
&lt;/h2&gt;

&lt;p&gt;If you have run a site behind a CDN yourself, you know the parts: a cache policy with three different time-to-live numbers, cache-control headers to get right per file type, an invalidation to request after every deploy, a script that does it, a wait of several minutes, and a second cache in the browser that no invalidation reaches. Most teams get one of these wrong at some point. Either the CDN keeps a stale site for a day and nobody knows why, or the caches are turned off in frustration and the site is slow for everyone, or a one-year header goes out on a file that then needs an urgent fix.&lt;/p&gt;

&lt;p&gt;Skapi's web hosting takes the four decisions that come out of that experience and makes them the only way it works. The CDN keeps your files for a day, so the site is fast everywhere. Your own upload clears it, about a minute later, so the site is fresh without a button. Browsers keep nothing unless you say so, so the default is safe. And when you do say so, the page warns you, the pages themselves stay out of it, and &lt;code&gt;?v=&lt;/code&gt; gives you a clean way out. The two rows on the Web Hosting page are the entire surface: one setting to choose, one state to watch. There is no console to open, no header to write and no script to run after a deploy.&lt;/p&gt;

&lt;p&gt;That is the shape of everything we build. The complexity of backend infrastructure is real, and it is ours to carry, not yours. You upload the files and build the site; the caching, the refreshing and the edge cases are taken care of, the same way the database needs no schema and the API needs no server. Your job is the thing you are making.&lt;/p&gt;

&lt;h2&gt;
  
  
  If you still see the old version
&lt;/h2&gt;

&lt;p&gt;A short checklist, in the order that usually finds it.&lt;/p&gt;

&lt;ol&gt;
&lt;li&gt;Look at the &lt;strong&gt;CDN cache&lt;/strong&gt; row. If it reads &lt;strong&gt;Updating&lt;/strong&gt;, the refresh is running; give it a minute. If it reads &lt;strong&gt;Live&lt;/strong&gt; and the change is older than a minute or two, press &lt;strong&gt;[Refresh now]&lt;/strong&gt; and tell us, because that should not happen.&lt;/li&gt;
&lt;li&gt;Check your own browser. The file you are looking at may be one your browser downloaded earlier under a longer setting. A reload fetches the page again, and the page fetches whatever it names; if the file's name has not changed and its allowance has not run out, your browser keeps its copy, by design. Change the address (&lt;code&gt;?v=2&lt;/code&gt;) or wait it out.&lt;/li&gt;
&lt;li&gt;Check the file's own value in the &lt;strong&gt;Browser caching&lt;/strong&gt; column of the file list. A file uploaded before you changed the setting still carries the old one. Upload it again.&lt;/li&gt;
&lt;li&gt;Remember that pages are never kept. If a page itself looks old, it is the CDN, and step 1 applies.&lt;/li&gt;
&lt;/ol&gt;

&lt;p&gt;The documentation has the details: &lt;a href="https://docs.skapi.com/hosting/hosting.html#the-cdn-cache" rel="noopener noreferrer"&gt;The CDN cache&lt;/a&gt; and &lt;a href="https://docs.skapi.com/hosting/hosting.html#browser-caching" rel="noopener noreferrer"&gt;Browser caching&lt;/a&gt; on the Web Hosting page.&lt;/p&gt;

</description>
      <category>webdev</category>
      <category>tutorial</category>
      <category>backend</category>
      <category>programming</category>
    </item>
    <item>
      <title>Building Stillwater Lake: a multiplayer 3D game in one HTML file, with Claude and Skapi</title>
      <dc:creator>Skapi</dc:creator>
      <pubDate>Mon, 09 Nov 2026 07:47:36 +0000</pubDate>
      <link>https://dev.to/skapi_api/building-stillwater-lake-a-multiplayer-3d-game-in-one-html-file-with-claude-and-skapi-4fhk</link>
      <guid>https://dev.to/skapi_api/building-stillwater-lake-a-multiplayer-3d-game-in-one-html-file-with-claude-and-skapi-4fhk</guid>
      <description>&lt;h2&gt;
  
  
  Building Stillwater Lake: a multiplayer 3D game in one HTML file, with Claude and Skapi
&lt;/h2&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fc29qrtr2pzdkar1pqtx9.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fc29qrtr2pzdkar1pqtx9.png" alt="Golden hour at Stillwater Lake: Hana casting from the dock, Mina fishing, friends dancing on the shore" width="800" height="420"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Play it:&lt;/strong&gt; &lt;a href="https://stillwater.skapi.com" rel="noopener noreferrer"&gt;stillwater.skapi.com&lt;/a&gt; · &lt;strong&gt;Source:&lt;/strong&gt; &lt;a href="https://github.com/broadwayinc/stillwater-lake-game" rel="noopener noreferrer"&gt;github.com/broadwayinc/stillwater-lake-game&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;Stillwater Lake is a quiet 3D lake you can walk around with whoever else is there. You cast a line, gather branches and stones, light a campfire, grill your catch, and sit down to eat a fish someone else cooked. Every fish you eat earns an appreciation dance. You can say a kind word to the lake, take a photo and share it, swim out to the middle, and wander over the ridge until the world loops you back to the same shore. The light follows your own clock, so it's dusk at the lake when it's dusk where you are.&lt;/p&gt;

&lt;p&gt;It runs on phones and computers, in Safari and Chrome, and even inside the Threads and Instagram in-app browsers. It has live multiplayer, guest accounts that need no sign-up, and its own web address.&lt;/p&gt;

&lt;p&gt;And here's the part that still feels unreal: &lt;strong&gt;the whole game is one HTML file, and it's smaller than your Instagram photo.&lt;/strong&gt; About 158 KB. A single photo on your phone is usually bigger. There's no app to install, no project folder with a hundred files and no server to run. Just one file.&lt;/p&gt;

&lt;p&gt;It was built almost entirely hands off. The developer described what they wanted in plain sentences in claude.ai, the same chat app people use every day. Claude wrote the game, showed it running right in the chat, fixed what didn't feel right, connected it to Skapi for multiplayer and accounts, and put it online at its own address. Nobody opened a terminal, set up a server or wrote a deployment script.&lt;/p&gt;




&lt;h2&gt;
  
  
  Software is becoming a document
&lt;/h2&gt;

&lt;p&gt;For decades, "building software" meant a toolchain: a framework, a package manager, a build pipeline, a backend, a database, a deployment process and a hosting account. Each piece needed its own expertise.&lt;/p&gt;

&lt;p&gt;Stillwater Lake is a single document, no different in kind from a Word file or a PDF. It holds everything it needs and opens anywhere. You can email it, put it in a folder or attach it to a chat. Skapi gives that file a backend (accounts, live multiplayer and hosting) without the file needing a server of its own. The AI gives anyone the ability to write it.&lt;/p&gt;

&lt;p&gt;That is the shift worth paying attention to. &lt;strong&gt;If you can describe it, you can now ship it.&lt;/strong&gt; The barrier is no longer knowing how to set up a stack. It's knowing what you want to make.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fslnwa0nlpny4mhga39en.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fslnwa0nlpny4mhga39en.png" alt="Friends gathered by the campfire. One is eating a grilled fish, one is sharing a kind word." width="800" height="500"&gt;&lt;/a&gt;&lt;/p&gt;




&lt;h2&gt;
  
  
  What Skapi gives the game
&lt;/h2&gt;

&lt;p&gt;Skapi is a backend you don't have to build: accounts, a database, file storage, live messaging and website hosting, all ready to use. Stillwater Lake uses three of them, and Claude wired up all three.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;1. Accounts with no sign-up.&lt;/strong&gt; Every visitor gets a guest account the moment they walk in. There's no e-mail, no password and no form. If someone leaves the tab for a long time and their guest account expires, the game notices, welcomes them back and gives them a fresh one.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;2. Live multiplayer.&lt;/strong&gt; Everyone at the lake is connected to the same live channel. A few times a second, each player tells everyone else where they are, what they're wearing and what they're doing, along with their campfires and the fish on their grill. Catches, kind words and dances show up for everyone the moment they happen.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;3. Its own web address.&lt;/strong&gt; The game lives at &lt;code&gt;stillwater.skapi.com&lt;/code&gt;. Claude registered the address, uploaded the file and put every new version online itself.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fljsyxpcic7ued223stxv.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fljsyxpcic7ued223stxv.png" alt="Two anglers on the dock. Sora's line is already out." width="800" height="500"&gt;&lt;/a&gt;&lt;/p&gt;




&lt;h2&gt;
  
  
  How the development was set up
&lt;/h2&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fz4gjheceq5kvq4sp1w7a.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fz4gjheceq5kvq4sp1w7a.png" alt="How the pieces fit together: you ask in plain words; inside claude.ai, Claude writes the HTML and uses the Skapi MCP connector to wire it to Skapi and upload it; your Skapi project hosts the page and runs the backend" width="800" height="466"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;There are two players inside claude.ai, and it helps to know who does what.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Claude writes the game.&lt;/strong&gt; Every line of the HTML, the 3D world, the controls and the animations come from Claude, and it shows you each version as a playable preview right in the chat. That includes the small piece of Skapi code inside the file that signs players in as guests and connects everyone to the same live channel.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;The Skapi MCP connector does the Skapi work.&lt;/strong&gt; MCP (Model Context Protocol) is a standard way to give an AI assistant extra tools. Once you connect Skapi's MCP server to claude.ai, Claude gains tools that act on your Skapi project as you. It reads the Skapi docs through the connector, so the backend code it writes into the HTML matches how Skapi actually works. It registers your subdomain, uploads the finished HTML to Skapi's web hosting and refreshes it when you ask for changes. It can also look at your project's users and data and change project settings.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Your Skapi project is the backend and the home.&lt;/strong&gt; It serves the page at your address and runs the guest accounts and live multiplayer. There's no server of your own in between: when someone opens the link, the Skapi code inside the HTML talks to your project directly.&lt;/p&gt;

&lt;p&gt;So you talk, Claude writes, and the connector ships. You never touch a terminal, an upload form or a hosting dashboard.&lt;/p&gt;

&lt;p&gt;This is the whole setup. It follows the &lt;a href="https://docs.skapi.com/introduction/getting-started.html" rel="noopener noreferrer"&gt;getting-started guide&lt;/a&gt; in the Skapi docs, using claude.ai in your browser or the Claude desktop app.&lt;/p&gt;

&lt;h3&gt;
  
  
  Step 1. Create a Skapi project
&lt;/h3&gt;

&lt;p&gt;Sign up at &lt;a href="https://www.skapi.com" rel="noopener noreferrer"&gt;skapi.com&lt;/a&gt;, then create a project at &lt;a href="https://www.skapi.com/new-project" rel="noopener noreferrer"&gt;skapi.com/new-project&lt;/a&gt;. Give it a name, pick a region and choose a plan.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Ffiq938cktjw09gxpe4v7.jpg" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Ffiq938cktjw09gxpe4v7.jpg" alt="Creating a new project on skapi.com: a name, an optional description and a region" width="800" height="600"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;Your new project opens on its own page. The &lt;strong&gt;Information&lt;/strong&gt; card shows your &lt;strong&gt;Project ID&lt;/strong&gt;; click it to copy. This tells Claude which project to build on.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fk264zqlzu5s5cm1fqvu3.jpg" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fk264zqlzu5s5cm1fqvu3.jpg" alt="The Stillwater Lake project page. The Project ID is on the Information card." width="800" height="560"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;h3&gt;
  
  
  Step 2. Connect Skapi to Claude
&lt;/h3&gt;

&lt;p&gt;This is the step that makes the whole thing hands off. The Skapi MCP server lets Claude work on your project for you: set up your web address, upload your game, put new versions online, look at your users and data, and change project settings. It signs in as you, so it only touches your own projects.&lt;/p&gt;

&lt;p&gt;You need one address. It's on the &lt;strong&gt;For AI Agents&lt;/strong&gt; card on your project page:&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fiaxtkonsgtfrd2hvbjb4.jpg" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fiaxtkonsgtfrd2hvbjb4.jpg" alt="The MCP endpoint on the For AI Agents card: https://mcp.broadwayinc.computer" width="800" height="262"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;Then, in claude.ai:&lt;/p&gt;

&lt;ol&gt;
&lt;li&gt;
&lt;p&gt;Open &lt;strong&gt;Customize&lt;/strong&gt;, go to &lt;strong&gt;Connectors&lt;/strong&gt;, click &lt;strong&gt;+ Add&lt;/strong&gt; and choose &lt;strong&gt;Add custom connector&lt;/strong&gt;.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fw5675c6yd06dsoy4jz3y.jpg" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fw5675c6yd06dsoy4jz3y.jpg" alt="The Connectors page in claude.ai, with the Add menu open" width="800" height="560"&gt;&lt;/a&gt;&lt;/p&gt;
&lt;/li&gt;
&lt;li&gt;
&lt;p&gt;Name it &lt;em&gt;Skapi&lt;/em&gt;, paste &lt;code&gt;https://mcp.broadwayinc.computer&lt;/code&gt; as the MCP server URL, and click &lt;strong&gt;Continue&lt;/strong&gt;.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fpbvdd74uf12uyatkyxay.jpg" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fpbvdd74uf12uyatkyxay.jpg" alt="The Add custom connector form, filled in for Skapi" width="800" height="560"&gt;&lt;/a&gt;&lt;/p&gt;
&lt;/li&gt;
&lt;li&gt;
&lt;p&gt;Sign in with your Skapi account when asked. There's no key or secret to copy anywhere. Skapi now shows up in your connectors with a check mark.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2F03karx1s87vcwz4augpo.jpg" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2F03karx1s87vcwz4augpo.jpg" alt="Skapi listed under Yours, connected" width="800" height="560"&gt;&lt;/a&gt;&lt;/p&gt;
&lt;/li&gt;
&lt;/ol&gt;

&lt;p&gt;On Team and Enterprise plans, an organization owner adds the connector once under &lt;strong&gt;Organization settings › Connectors&lt;/strong&gt;, and then each member just clicks &lt;strong&gt;Connect&lt;/strong&gt;.&lt;/p&gt;

&lt;p&gt;That's it. Claude can now see your Skapi projects and build on them. You don't need to download or attach anything else.&lt;/p&gt;

&lt;h3&gt;
  
  
  Step 3. Describe what you want
&lt;/h3&gt;

&lt;p&gt;Open a new chat in claude.ai and just start typing. Tell Claude which project to use, then describe your idea in your own words.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fq7bcasehs3owiy3ggd98.jpg" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fq7bcasehs3owiy3ggd98.jpg" alt="A new chat in claude.ai with the Stillwater Lake prompt being typed: &amp;quot;Use my Skapi project &lt;Project ID&gt;&amp;quot;, then the game description" width="800" height="264"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;The Stillwater Lake prompt started like this:&lt;/p&gt;

&lt;blockquote&gt;
&lt;p&gt;Use my Skapi project "&amp;lt;Project ID&amp;gt;".&lt;/p&gt;

&lt;p&gt;Build "Stillwater Lake": a quiet, shared 3D lake that anyone can walk into from a phone or a computer, with no sign-up. Everyone at the lake sees everyone else in real time. Show me the game as a playable preview here in the chat after each step…&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;The world.&lt;/strong&gt; A round lake in a basin, with a sandy shore, grassy rolling hills, rocky mountains with snow on the peaks… A wooden dock reaching into the lake, with a glowing lantern… A base camp near the dock: a stone-ringed campfire…&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Fishing.&lt;/strong&gt; Near the water, Space (or a big round "Cast" button on phones) casts toward the middle of the lake… After 4 to 13 seconds it bites…&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Camp life.&lt;/strong&gt; With 3 branches and 2 stones, build a campfire anywhere on land… Anyone at the lake can eat your grilled fish… Every fish you eat banks one appreciation dance: a cha-cha after the 1996 "Dancing Baby"…&lt;/p&gt;
&lt;/blockquote&gt;

&lt;p&gt;You don't have to get it all right the first time. Play the preview, then just say what you'd like changed, the way you'd talk to a friend who's helping you. &lt;strong&gt;&lt;a href="https://cdn.broadwayinc.com/articles/stillwater-lake-prompt.md" rel="noopener noreferrer"&gt;The full Stillwater Lake prompt&lt;/a&gt;&lt;/strong&gt; is further down if you want to build your own copy in one go.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fhw79slivy21mhabzwj2i.jpg" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fhw79slivy21mhabzwj2i.jpg" alt="Claude in claude.ai: the conversation on the left, the game running right in the chat on the right" width="800" height="506"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Flrloipy64cbd0gvjmz9z.jpg" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Flrloipy64cbd0gvjmz9z.jpg" alt="The same game opened full size inside claude.ai, already playable before it ever went online" width="800" height="506"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;h3&gt;
  
  
  Step 4. Ask Claude to put it online
&lt;/h3&gt;

&lt;p&gt;Every Skapi project can host a website at &lt;code&gt;https://&amp;lt;your-name&amp;gt;.skapi.com&lt;/code&gt;. &lt;strong&gt;You need your own address&lt;/strong&gt;, since &lt;code&gt;stillwater&lt;/code&gt; is already taken by this game. You don't have to set it up yourself. Just ask Claude in the chat:&lt;/p&gt;

&lt;blockquote&gt;
&lt;p&gt;Register the subdomain "my-lake" on my Skapi project and put the game online there.&lt;/p&gt;
&lt;/blockquote&gt;

&lt;p&gt;Through the MCP connector, Claude registers the address, uploads your game and gives you the link. When you ask for changes later, Claude puts the new version online the same way. One thing stays with you on purpose: changing or removing an address once it's registered is done by hand on the Skapi dashboard.&lt;/p&gt;




&lt;h2&gt;
  
  
  What "hands off" actually looked like
&lt;/h2&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Faifewhnpidixem8l2t5a.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Faifewhnpidixem8l2t5a.png" alt="The game on a phone: Mina doing her appreciation dance by the lake" width="800" height="1628"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;The developer never left the chat window. Here's what Claude handled on its side of the conversation:&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;It showed the game running, right in the chat.&lt;/strong&gt; Every version appeared as a live, playable preview inside claude.ai. You could walk around, cast a line and try the new feature before anything went public. Feedback was as simple as playing it and saying what felt off: "the dance feels glitchy", "both hands should touch the fish", "make the name box as tall as the title card".&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fkr8njli97sg4mlbdtbym.jpg" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fkr8njli97sg4mlbdtbym.jpg" alt="Claude's work steps in the chat: changing the game, running its tests, checking its own screenshots" width="800" height="506"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;It tested its own work.&lt;/strong&gt; Before saying something was done, Claude opened the game in a browser of its own at phone and desktop sizes. It walked the character through campfires, grilled and ate fish, pretended to be a second player joining, and checked what happens when a guest account expires. When something failed, it fixed it and tried again.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;It solved the hard problem by itself.&lt;/strong&gt; Players coming back after a while couldn't reconnect. Claude dug into how Skapi's guest accounts work and found the cause: a phone pauses background tabs, and after half an hour of silence the guest account is cleaned up. Claude built the fix: notice the lost account, sign out cleanly, and show the welcome screen so walking back in starts fresh.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;It put every version online.&lt;/strong&gt; When asked to ship, Claude uploaded the game through the Skapi connector, checked that every byte arrived intact, refreshed the site and confirmed it was live. It also kept a backup copy of the previous version, so there was always a way back.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fhnn4uml8by9hk30c8n05.jpg" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fhnn4uml8by9hk30c8n05.jpg" alt="Claude putting the game online through the Skapi connector: the whole game goes up as one file" width="800" height="506"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;It took care of things nobody asked for.&lt;/strong&gt; It made the preview inside claude.ai multiplayer too. It reconnects quietly when the network drops. People who just arrived don't see a replay of old events. Campfires burn for exactly as long as their builder is at the lake. And in the Threads and Instagram in-app browsers, where photos can't be saved, it offers to open the game in your normal browser.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Folgn0cxx3qd9hswt7ggt.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Folgn0cxx3qd9hswt7ggt.png" alt="Customizing your angler on a phone: name, shirt, pants, cap and shoes" width="800" height="1628"&gt;&lt;/a&gt;&lt;/p&gt;




&lt;h2&gt;
  
  
  Take a photo, share the lake
&lt;/h2&gt;

&lt;p&gt;Press &lt;strong&gt;P&lt;/strong&gt; or tap the camera button, and the game captures the scene, adds a small caption with your name, and opens a sheet with &lt;strong&gt;Share&lt;/strong&gt; and &lt;strong&gt;Save&lt;/strong&gt;. On phones, Share goes straight to your phone's share menu: Messages, Instagram, AirDrop. Inside an in-app browser, the camera button explains why photos need a real browser and offers to open one.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2F9znqpglgucc86x705vbn.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2F9znqpglgucc86x705vbn.png" alt="The photo sheet: a snapshot from the lake, ready to share" width="800" height="500"&gt;&lt;/a&gt;&lt;/p&gt;




&lt;h2&gt;
  
  
  The takeaway
&lt;/h2&gt;

&lt;p&gt;Stillwater Lake has 3D graphics, a day and night cycle, swimming, fishing, cooking, a shared economy of grilled fish, and a dance borrowed from 1996. It has live multiplayer, guest accounts with no sign-up, touch controls, photo sharing and its own address on the web. And the whole thing is one file, smaller than a photo on your phone.&lt;/p&gt;

&lt;p&gt;Nobody wrote it line by line. Someone described it in a chat, Claude built it, and Skapi gave it a backend and a home, with the Skapi connector doing the setup and the shipping. The only "technical" step was pasting one address into Claude's connector settings.&lt;/p&gt;

&lt;p&gt;If you have an idea for a small shared place on the internet, you can build it this afternoon:&lt;/p&gt;

&lt;ol&gt;
&lt;li&gt;
&lt;a href="https://www.skapi.com/new-project" rel="noopener noreferrer"&gt;Create a Skapi project&lt;/a&gt; and copy its Project ID.&lt;/li&gt;
&lt;li&gt;Add &lt;code&gt;https://mcp.broadwayinc.computer&lt;/code&gt; as a custom connector in claude.ai and sign in.&lt;/li&gt;
&lt;li&gt;Tell Claude your Project ID and describe what you want, or paste &lt;a href="https://cdn.broadwayinc.com/articles/stillwater-lake-prompt.md" rel="noopener noreferrer"&gt;the Stillwater Lake prompt&lt;/a&gt;.&lt;/li&gt;
&lt;li&gt;Play the preview in the chat and say what to change.&lt;/li&gt;
&lt;li&gt;Ask Claude to register your own address and put it online.&lt;/li&gt;
&lt;/ol&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2F5xpogvqmhtfyo0q3nkqd.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2F5xpogvqmhtfyo0q3nkqd.png" alt="Night at the lake, by the campfire" width="800" height="500"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;&lt;em&gt;Stillwater Lake is open source: &lt;a href="https://github.com/broadwayinc/stillwater-lake-game" rel="noopener noreferrer"&gt;github.com/broadwayinc/stillwater-lake-game&lt;/a&gt;. Built on &lt;a href="https://skapi.com" rel="noopener noreferrer"&gt;skapi.com&lt;/a&gt;.&lt;/em&gt;&lt;/p&gt;

</description>
      <category>gamedev</category>
      <category>gamechallenge</category>
      <category>html</category>
      <category>tutorial</category>
    </item>
    <item>
      <title>A video call app that survives a bad connection</title>
      <dc:creator>Skapi</dc:creator>
      <pubDate>Wed, 04 Nov 2026 08:08:53 +0000</pubDate>
      <link>https://dev.to/skapi_api/a-video-call-app-that-survives-a-bad-connection-4oha</link>
      <guid>https://dev.to/skapi_api/a-video-call-app-that-survives-a-bad-connection-4oha</guid>
      <description>&lt;h2&gt;
  
  
  A video call that survives a bad connection
&lt;/h2&gt;

&lt;p&gt;Take a video call on a train. The connection comes and goes, and for a few weeks the &lt;a href="https://videocall.skapi.com/" rel="noopener noreferrer"&gt;video call app&lt;/a&gt; we published in October handled that the way most first versions do: the moment the browser reported the connection as dropped, the call was over, both of you were back in the waiting room, and a new call started at once. On a weak connection that meant being thrown out and pulled back in every few seconds, which is worse than a frozen picture. We fixed it, and the fix is not in the app alone. Most of it is in skapi-js 2.2.21, so every app built on Skapi's realtime connection and WebRTC calls gets it. This article explains what changed, with pictures, and ends with what it did for the video call app and what it means for yours.&lt;/p&gt;

&lt;h2&gt;
  
  
  Two connections, two jobs
&lt;/h2&gt;

&lt;p&gt;A call made with Skapi runs on two connections, and it helps to keep them apart from the start.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fna1bimk95twluo2qr9pz.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fna1bimk95twluo2qr9pz.png" alt="two connections" width="800" height="345"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;&lt;em&gt;Two connections with two jobs. The thin one is for finding each other; the thick one is the call.&lt;/em&gt;&lt;/p&gt;

&lt;p&gt;The thin one is the &lt;strong&gt;realtime connection&lt;/strong&gt; to Skapi, a WebSocket. It carries who is in the room and the handful of messages two browsers exchange to set a call up: the offer, the answer, the network addresses each side can be reached at. The thick one is &lt;strong&gt;the call itself&lt;/strong&gt;, a WebRTC connection that runs straight from one browser to the other. Video, voice, the chat messages and the files travel on it. Skapi is not in the middle of it, and once the call is up it does not even need the thin line, except to set up a change later: a screen share, a new path.&lt;/p&gt;

&lt;p&gt;A bad connection can break either of the two, or both. The improvements in 2.2.21 cover each.&lt;/p&gt;

&lt;h2&gt;
  
  
  What a drop used to do
&lt;/h2&gt;

&lt;p&gt;Here is the sequence that produced the train experience.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fo8x7hgiedvzi8fev8bv3.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fo8x7hgiedvzi8fev8bv3.png" alt="train experience" width="798" height="246"&gt;&lt;/a&gt;&lt;br&gt;
&lt;em&gt;The old sequence. A momentary state was treated as the end.&lt;/em&gt;&lt;/p&gt;

&lt;p&gt;A WebRTC connection reports its state to the page, and one of those states is &lt;code&gt;disconnected&lt;/code&gt;. It means "no packets for a few seconds". It is not the end of anything: browsers leave that state by themselves most of the time, when the packets start flowing again, and when they do not, a procedure called an ICE restart finds a new path, which is what a phone needs after moving from Wi-Fi to mobile data. The SDK used to close the connection the moment that state appeared. The app saw a closed call, did what it should for a closed call, and dialled again, because a lost call may be retried. On a connection that keeps blinking, that is a loop.&lt;/p&gt;
&lt;h2&gt;
  
  
  What happens now
&lt;/h2&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fa9kasq5e6l2rr2igkpl5.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fa9kasq5e6l2rr2igkpl5.png" alt="the new secuence" width="800" height="378"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;&lt;em&gt;The new sequence, in its two outcomes.&lt;/em&gt;&lt;/p&gt;

&lt;p&gt;The SDK no longer closes a connection that says &lt;code&gt;disconnected&lt;/code&gt;. It waits three seconds, because the browser often brings the connection back on its own within that time, then sends an ICE restart so the two browsers look for a path again. If the browser reports &lt;code&gt;failed&lt;/code&gt;, one more restart goes out at once. Only when nothing has helped within fifteen seconds of the first report is the call given up, and then it is given up on both sides at the same moment, so neither page is left talking to nobody.&lt;/p&gt;

&lt;p&gt;The page is told at each step, through the same callback it already uses for the call, with three events:&lt;/p&gt;

&lt;div class="table-wrapper-paragraph"&gt;&lt;table&gt;
&lt;thead&gt;
&lt;tr&gt;
&lt;th&gt;Event&lt;/th&gt;
&lt;th&gt;When&lt;/th&gt;
&lt;/tr&gt;
&lt;/thead&gt;
&lt;tbody&gt;
&lt;tr&gt;
&lt;td&gt;&lt;code&gt;reconnecting&lt;/code&gt;&lt;/td&gt;
&lt;td&gt;The connection dropped. The call is still open; show it, and keep the picture on the screen.&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;&lt;code&gt;reconnected&lt;/code&gt;&lt;/td&gt;
&lt;td&gt;The connection is back. Nothing else changed: the tracks, the data channels and the files in flight are the same.&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;&lt;code&gt;lost&lt;/code&gt;&lt;/td&gt;
&lt;td&gt;The connection could not be brought back. The call is closed right after.&lt;/td&gt;
&lt;/tr&gt;
&lt;/tbody&gt;
&lt;/table&gt;&lt;/div&gt;

&lt;p&gt;The one that matters most for an app is the third, because a lost call and a hang-up look the same from the outside: both end with a &lt;code&gt;closed&lt;/code&gt; state. The &lt;code&gt;lost&lt;/code&gt; event before the close is how a page tells them apart, and it is why the video call app can dial again after a loss and show a button after a hang-up.&lt;/p&gt;
&lt;h2&gt;
  
  
  The thin line reopens by itself
&lt;/h2&gt;

&lt;p&gt;The realtime connection is a WebSocket, and a WebSocket on a phone closes whenever the network goes away for more than a moment. The SDK has always reconnected it, but it used to stop after a few tries, and an attempt that failed because the network was still down was the last one. A page that came out of a tunnel after a minute had no realtime connection and no way to get one back short of a reload.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2F8i2abqx8q39rt07y3464.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2F8i2abqx8q39rt07y3464.png" alt="o realtime connection" width="800" height="239"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;&lt;em&gt;The realtime connection comes back by itself, at a pace that does not hammer anything.&lt;/em&gt;&lt;/p&gt;

&lt;p&gt;Now the attempts go on: the first three seconds after the close, then six, twelve and twenty-four seconds apart, then every thirty seconds, for as long as the page is open and the user is signed in. A page that is brought back to the foreground tries at once. The group the user was in is joined again on the first success, and the page hears &lt;code&gt;reconnect&lt;/code&gt; before each attempt and &lt;code&gt;success&lt;/code&gt; when one works, so it can say "Reconnecting..." and clear it. A call that is up goes on through all of this; it does not depend on the thin line.&lt;/p&gt;
&lt;h2&gt;
  
  
  A new address, the same call
&lt;/h2&gt;

&lt;p&gt;There is a subtlety here that is worth a picture, because it is the part that makes the two improvements work together.&lt;/p&gt;

&lt;p&gt;Every realtime connection has a connection id, and a call is addressed by the other side's id: when Ana's browser sends Ben's browser the offer, the answer or a new network address, it sends it to Ben's connection id. When Ben's realtime connection drops and reopens, the server gives it a new id. Ana does not know that. Her next message for Ben, an ICE restart for example, would go to an id that no longer exists, and the call that was about to be repaired would be lost after all.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fdlvbymwh0j3e6q02obvy.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fdlvbymwh0j3e6q02obvy.png" alt="call keeps its name" width="799" height="340"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;&lt;em&gt;The call keeps its name across a change of connection id.&lt;/em&gt;&lt;/p&gt;

&lt;p&gt;So every signal of a call now names the call, with an id both sides share from the first offer. When Ben's connection reopens, his browser sends an empty signal to Ana: "call k7p2, still here", and it arrives from his new connection id. Ana's browser has never seen that id, but it has the call, so it learns that Ben is at the new id now and writes there from then on. The app hears a &lt;code&gt;cidchange&lt;/code&gt; event with the new id and the old one. The call itself never noticed; it was running between the browsers the whole time.&lt;/p&gt;
&lt;h2&gt;
  
  
  Signed in through it all
&lt;/h2&gt;

&lt;p&gt;One more piece, and it is not about connections at all. A Skapi session is kept alive by refreshing its sign-in token in the background, and a refresh needs the network. Before 2.2.21 a refresh that failed for any reason signed the user out, and "any reason" included "the request never reached the server". A long enough outage at the wrong moment ended with the user on the sign-in page, and with the realtime reconnect stopped, since there was no session left to reconnect with. For the video call app, whose accounts are anonymous and temporary, it meant a new account and a room that no longer knew the person.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fkojul8ej6ymd77388rep.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fkojul8ej6ymd77388rep.png" alt="answer from the server ends a session" width="800" height="281"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;&lt;em&gt;Only an answer from the server ends a session now.&lt;/em&gt;&lt;/p&gt;

&lt;p&gt;The two cases are told apart now. A refresh the server rejects, because the token is expired or revoked or the account is gone, signs the user out, as it should. A refresh that got no answer keeps the session and the token as they are; the request that needed it fails with the code &lt;code&gt;NETWORK_ERROR&lt;/code&gt;, the way every request fails without network, and the next request refreshes again. Mail and banking apps on a phone have worked this way for years, and it is the only sensible way: a tunnel is not a reason to sign anyone out.&lt;/p&gt;
&lt;h2&gt;
  
  
  What it did for the video call app
&lt;/h2&gt;

&lt;p&gt;The app is open source, and it is where all of this was tried first. Four things changed on the screen.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;The call stays on the screen.&lt;/strong&gt; When the connection drops, a "Reconnecting..." pill appears over the picture, the status dot pulses, the picture freezes, and a line in the chat says what is happening. Most of the time the next line is "Connected again." and nothing else happened. When the call is really lost, both pages go back to the waiting room, it says so, and a new call starts by itself.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;The browser asks for the camera once.&lt;/strong&gt; This one was a separate complaint, and the cause was separate too: the app opened the camera for the preview in the waiting room, stopped it when the call started, and the SDK opened it again for the call. Two captures, and in Firefox and Safari two permission prompts.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fr9rvxlpsiv2h3mih7c9z.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fr9rvxlpsiv2h3mih7c9z.png" alt="One opening of the camera" width="799" height="310"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;&lt;em&gt;One opening of the camera for the whole visit.&lt;/em&gt;&lt;/p&gt;

&lt;p&gt;Now the camera and the microphone are opened once, for the preview, and stay open for the visit. Each call sends a copy of those tracks, made with &lt;code&gt;track.clone()&lt;/code&gt;, which never prompts; the SDK stops the copies when the call ends and the preview keeps the originals for the next one. This needed one small addition to the SDK: &lt;code&gt;media.audio&lt;/code&gt; and &lt;code&gt;media.video&lt;/code&gt; accept a &lt;code&gt;MediaStreamTrack&lt;/code&gt; now, sent as the microphone or the camera without a new capture.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;A pause before dialling again.&lt;/strong&gt; When a call is lost, the waiting room waits a second before the next one, then two, four and eight seconds for losses in a row. A call that connects starts the count over. It is the difference between a page that tries again sensibly and one that hammers a dead connection.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fkz7h0beta73ekqssyrb5.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fkz7h0beta73ekqssyrb5.png" alt="The pauses after lost calls" width="799" height="197"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;&lt;em&gt;The pauses after lost calls. Eight seconds is the most it ever waits.&lt;/em&gt;&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;The chat and the board survive.&lt;/strong&gt; A lost call keeps the chat log and the drawing, and the next call with the same person goes on where it stopped: the log says "Mina is back." instead of starting empty, and the board opens again with the drawing on it. Only a hang-up on purpose starts a fresh call.&lt;/p&gt;

&lt;p&gt;There is one more thing the app learned to do, and it is the kind of detail that only shows up on a real bad network. The room lists connections, not people, and the service learns that a connection died only when it tries to send to it. So for a while after someone's network dropped, the room can list that person twice: the dead connection and the new one. The app used to count two and call the room full. It counts people now, calls the newest connection of each, and when the service answers that a connection is gone, the entry goes and the call is dialled again to the one that is alive.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto/https%3A%2F%2Fcdn.broadwayinc.com%2Farticles%2Fimg%2Fvideo-call-call.webp" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto/https%3A%2F%2Fcdn.broadwayinc.com%2Farticles%2Fimg%2Fvideo-call-call.webp" alt="A call in progress in the video call app: the other person on the stage, the two small pictures below it, and the chat open beside the call" width="800" height="500"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;&lt;em&gt;The app, at &lt;a href="https://videocall.skapi.com/" rel="noopener noreferrer"&gt;videocall.skapi.com&lt;/a&gt;. Free, no sign-up, and the source is on &lt;a href="https://github.com/broadwayinc/skapi-video-call" rel="noopener noreferrer"&gt;GitHub&lt;/a&gt;.&lt;/em&gt;&lt;/p&gt;
&lt;h2&gt;
  
  
  In your own app
&lt;/h2&gt;

&lt;p&gt;Everything above comes with skapi-js 2.2.21 and needs no change on the server. If your app already ends a call only on the &lt;code&gt;closed&lt;/code&gt; state, it gets the recovery for free and can ignore the new events. If it ends the call on &lt;code&gt;disconnected&lt;/code&gt;, as the earlier docs example did, it should stop: that state is no longer the end, and a page that tears itself down on it would be showing "call ended" over a call that is about to come back. The pattern is short:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight javascript"&gt;&lt;code&gt;&lt;span class="kd"&gt;function&lt;/span&gt; &lt;span class="nf"&gt;RTCCallback&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="nx"&gt;e&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt; &lt;span class="p"&gt;{&lt;/span&gt;
    &lt;span class="k"&gt;switch &lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="nx"&gt;e&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nx"&gt;type&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt; &lt;span class="p"&gt;{&lt;/span&gt;
        &lt;span class="k"&gt;case&lt;/span&gt; &lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="s1"&gt;reconnecting&lt;/span&gt;&lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt;
            &lt;span class="nx"&gt;status&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nx"&gt;textContent&lt;/span&gt; &lt;span class="o"&gt;=&lt;/span&gt; &lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="s1"&gt;Reconnecting...&lt;/span&gt;&lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="p"&gt;;&lt;/span&gt;
            &lt;span class="k"&gt;break&lt;/span&gt;&lt;span class="p"&gt;;&lt;/span&gt;
        &lt;span class="k"&gt;case&lt;/span&gt; &lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="s1"&gt;reconnected&lt;/span&gt;&lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt;
            &lt;span class="nx"&gt;status&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nx"&gt;textContent&lt;/span&gt; &lt;span class="o"&gt;=&lt;/span&gt; &lt;span class="dl"&gt;''&lt;/span&gt;&lt;span class="p"&gt;;&lt;/span&gt;
            &lt;span class="k"&gt;break&lt;/span&gt;&lt;span class="p"&gt;;&lt;/span&gt;
        &lt;span class="k"&gt;case&lt;/span&gt; &lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="s1"&gt;lost&lt;/span&gt;&lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt;
            &lt;span class="nx"&gt;lost&lt;/span&gt; &lt;span class="o"&gt;=&lt;/span&gt; &lt;span class="kc"&gt;true&lt;/span&gt;&lt;span class="p"&gt;;&lt;/span&gt; &lt;span class="c1"&gt;// the closed that follows is not a hang-up&lt;/span&gt;
            &lt;span class="k"&gt;break&lt;/span&gt;&lt;span class="p"&gt;;&lt;/span&gt;
        &lt;span class="k"&gt;case&lt;/span&gt; &lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="s1"&gt;connectionstatechange&lt;/span&gt;&lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt;
            &lt;span class="k"&gt;if &lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="nx"&gt;e&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nx"&gt;state&lt;/span&gt; &lt;span class="o"&gt;===&lt;/span&gt; &lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="s1"&gt;closed&lt;/span&gt;&lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt; &lt;span class="nf"&gt;endCall&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="nx"&gt;lost&lt;/span&gt;&lt;span class="p"&gt;);&lt;/span&gt;
            &lt;span class="k"&gt;break&lt;/span&gt;&lt;span class="p"&gt;;&lt;/span&gt;
    &lt;span class="p"&gt;}&lt;/span&gt;
&lt;span class="p"&gt;}&lt;/span&gt;
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;And for the realtime connection, one message per attempt:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight javascript"&gt;&lt;code&gt;&lt;span class="kd"&gt;let&lt;/span&gt; &lt;span class="nx"&gt;RealtimeCallback&lt;/span&gt; &lt;span class="o"&gt;=&lt;/span&gt; &lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="nx"&gt;rt&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt; &lt;span class="o"&gt;=&amp;gt;&lt;/span&gt; &lt;span class="p"&gt;{&lt;/span&gt;
    &lt;span class="k"&gt;if &lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="nx"&gt;rt&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nx"&gt;type&lt;/span&gt; &lt;span class="o"&gt;===&lt;/span&gt; &lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="s1"&gt;reconnect&lt;/span&gt;&lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt; &lt;span class="nx"&gt;status&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nx"&gt;textContent&lt;/span&gt; &lt;span class="o"&gt;=&lt;/span&gt; &lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="s1"&gt;Connection lost. Reconnecting...&lt;/span&gt;&lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="p"&gt;;&lt;/span&gt;
    &lt;span class="k"&gt;if &lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="nx"&gt;rt&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nx"&gt;type&lt;/span&gt; &lt;span class="o"&gt;===&lt;/span&gt; &lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="s1"&gt;success&lt;/span&gt;&lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt; &lt;span class="nx"&gt;status&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nx"&gt;textContent&lt;/span&gt; &lt;span class="o"&gt;=&lt;/span&gt; &lt;span class="dl"&gt;''&lt;/span&gt;&lt;span class="p"&gt;;&lt;/span&gt;
&lt;span class="p"&gt;};&lt;/span&gt;
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;The details are in the documentation: &lt;a href="https://docs.skapi.com/realtime/webRTC.html#a-dropped-connection" rel="noopener noreferrer"&gt;A dropped connection&lt;/a&gt;, &lt;a href="https://docs.skapi.com/realtime/connecting.html#staying-connected" rel="noopener noreferrer"&gt;Staying connected&lt;/a&gt;, the &lt;code&gt;media&lt;/code&gt; option in &lt;a href="https://docs.skapi.com/realtime/webRTC.html#what-a-call-can-carry" rel="noopener noreferrer"&gt;What a call can carry&lt;/a&gt;, and the &lt;a href="https://docs.skapi.com/versionlog/versions.html" rel="noopener noreferrer"&gt;version history&lt;/a&gt; for 2.2.21.&lt;/p&gt;

&lt;p&gt;A connection that blinks should be a frozen picture for a few seconds, not a call that ends. That is what it is now, for the video call app and for every app that makes calls on Skapi.&lt;/p&gt;

</description>
      <category>productivity</category>
      <category>tutorial</category>
      <category>opensource</category>
      <category>api</category>
    </item>
    <item>
      <title>Open Source WebRTC Video Call App - use for free or build your own</title>
      <dc:creator>Skapi</dc:creator>
      <pubDate>Mon, 02 Nov 2026 08:59:12 +0000</pubDate>
      <link>https://dev.to/skapi_api/open-source-webrtc-video-call-app-use-for-free-or-build-your-own-36dj</link>
      <guid>https://dev.to/skapi_api/open-source-webrtc-video-call-app-use-for-free-or-build-your-own-36dj</guid>
      <description>&lt;h2&gt;
  
  
  Video Call: a full video call in a lightweight HTML page, open source on Skapi
&lt;/h2&gt;

&lt;p&gt;We built a video call. Not a product, a demonstration: an open source video call for two people that runs in the browser, with screen sharing, a shared drawing board, a chat that takes markdown, and file transfer. It is called Video Call, it is made of a lightweight HTML page and three small files with no build step, and the code is at &lt;a href="https://github.com/broadwayinc/skapi-video-call" rel="noopener noreferrer"&gt;github.com/broadwayinc/skapi-video-call&lt;/a&gt;. It is live at &lt;a href="https://videocall.skapi.com/" rel="noopener noreferrer"&gt;videocall.skapi.com&lt;/a&gt;, and anybody can use it there for free: open the page, send the link, talk. Anybody can also put it on their own website, customize it, or host it on their own Skapi project by changing one line: the project ID.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto/https%3A%2F%2Fcdn.broadwayinc.com%2Farticles%2Fimg%2Fvideo-call-call.webp" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto/https%3A%2F%2Fcdn.broadwayinc.com%2Farticles%2Fimg%2Fvideo-call-call.webp" alt="A call in progress: the other person on the stage, the two small pictures below it, and the chat open beside the call" width="800" height="500"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;&lt;em&gt;A call in Video Call. The large picture is the stage, the small ones are the two cameras, and the chat sits beside it.&lt;/em&gt;&lt;/p&gt;

&lt;p&gt;We wanted to show two things with it. The first is how far Skapi can take the RTC connection that is already built into every browser. The video, the voice, a shared screen, the drawing, the chat and the files all travel straight from one browser to the other over Skapi's RTC connection, with no server in between, and the whole application fits in one page. The second is a new feature of Skapi, &lt;a href="https://docs.skapi.com/authentication/anonymous-login.html" rel="noopener noreferrer"&gt;&lt;code&gt;anonymousLogin()&lt;/code&gt;&lt;/a&gt;, which is what lets anyone open the page and talk without creating an account. This article walks through what the app does, how it works, and how to make it yours.&lt;/p&gt;

&lt;h2&gt;
  
  
  What it does
&lt;/h2&gt;

&lt;p&gt;Open the page and you are in the waiting room: your camera preview, your name, and a link. Send the link to the person you want to talk to. They open it, check their own preview, and press Enter. The call starts by itself. Nobody signs up, nobody installs anything, and nobody has to accept a call: having the link is the invitation.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto/https%3A%2F%2Fcdn.broadwayinc.com%2Farticles%2Fimg%2Fvideo-call-waiting-room.webp" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto/https%3A%2F%2Fcdn.broadwayinc.com%2Farticles%2Fimg%2Fvideo-call-waiting-room.webp" alt="The waiting room: the camera preview with the microphone and camera buttons over it, the name field, the link with a Copy button, and the status line" width="800" height="500"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;&lt;em&gt;The waiting room. The link is the whole invitation.&lt;/em&gt;&lt;/p&gt;

&lt;p&gt;During the call, the large picture is the stage and the small pictures below it are everything in the call: your camera, the other person's camera, a shared screen, the drawing board. Click a small picture to put it on the stage. The buttons at the bottom mute your microphone, stop your camera, share your screen, open the drawing board, open the chat, and leave.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto/https%3A%2F%2Fcdn.broadwayinc.com%2Farticles%2Fimg%2Fvideo-call-screen-share.webp" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto/https%3A%2F%2Fcdn.broadwayinc.com%2Farticles%2Fimg%2Fvideo-call-screen-share.webp" alt="The other person's shared screen on the stage, a document in a word processor, with the chat open beside it" width="800" height="500"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;&lt;em&gt;Share a window, a tab or your whole screen. It goes to the other person's stage right away.&lt;/em&gt;&lt;/p&gt;

&lt;p&gt;The drawing board is a white board that both of you draw on at the same time, with six pens, three sizes, an eraser and a clear button. Everything you draw shows up on the other side as you draw it.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto/https%3A%2F%2Fcdn.broadwayinc.com%2Farticles%2Fimg%2Fvideo-call-board.webp" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto/https%3A%2F%2Fcdn.broadwayinc.com%2Farticles%2Fimg%2Fvideo-call-board.webp" alt="The drawing board on the stage with a red and a blue line on it, the tools floating at the top, and the chat beside it" width="800" height="500"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;&lt;em&gt;The shared drawing board. The board, the drawing and the tools live in the page; the strokes travel over the call.&lt;/em&gt;&lt;/p&gt;

&lt;p&gt;The chat takes markdown: bold, headings, lists, links, tables and code blocks with syntax colouring. Enter sends, Shift+Enter adds a line, Tab indents, like a code editor. Messages go straight to the other browser and are not kept anywhere; when the call ends, the chat is gone. A file sent from the paperclip goes over the call too, with a progress bar on both sides, and pictures show right in the chat.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto/https%3A%2F%2Fcdn.broadwayinc.com%2Farticles%2Fimg%2Fvideo-call-chat.webp" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto/https%3A%2F%2Fcdn.broadwayinc.com%2Farticles%2Fimg%2Fvideo-call-chat.webp" alt="The chat beside a call, with a highlighted JavaScript code block, a quoted line, links, a small table and a Python one-liner" width="800" height="500"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;&lt;em&gt;The chat renders markdown, and a code block named with its language is coloured.&lt;/em&gt;&lt;/p&gt;

&lt;p&gt;Everything works the same on a phone, laid out for a small screen.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto/https%3A%2F%2Fcdn.broadwayinc.com%2Farticles%2Fimg%2Fvideo-call-phone-call.webp" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto/https%3A%2F%2Fcdn.broadwayinc.com%2Farticles%2Fimg%2Fvideo-call-phone-call.webp" alt="A call on a phone: the other person on the stage, the two small pictures, and the icon-only control bar" width="520" height="1125"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;&lt;em&gt;On a phone the chat covers the call while it is open, and the buttons show icons only.&lt;/em&gt;&lt;/p&gt;

&lt;p&gt;And a word on privacy, because it is the point of the design. When you open the page you get a temporary account that is deleted about half an hour after you leave. Your video, voice, messages, drawings and files travel directly between the two browsers and are not stored anywhere. The service behind the app only helps the two browsers find each other.&lt;/p&gt;

&lt;h2&gt;
  
  
  How it works
&lt;/h2&gt;

&lt;p&gt;Three things happen when a call comes about: each browser signs in, the two meet in a room, and then they connect to each other directly. Skapi is in the first two. It is not in the third.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2F02i7hgf834mgf5drsk5q.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2F02i7hgf834mgf5drsk5q.png" alt="Skapi video call app how it works" width="800" height="386"&gt;&lt;/a&gt;&lt;br&gt;
&lt;em&gt;Skapi does the sign-in, keeps the room, and relays the handshake. The call itself is a WebRTC connection between the two browsers.&lt;/em&gt;&lt;/p&gt;

&lt;p&gt;Here is the same thing as a sequence, the way the code does it.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fsokcz7gdq52mr8e5oddv.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fsokcz7gdq52mr8e5oddv.png" alt="Video app code" width="799" height="630"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;&lt;em&gt;Every step until the handshake goes through Skapi's realtime connection. The connection the handshake sets up does not.&lt;/em&gt;&lt;/p&gt;

&lt;ol&gt;
&lt;li&gt;The host opens the page: a temporary account, no sign-up.&lt;/li&gt;
&lt;li&gt;A room id goes into the address; the host joins its realtime group.&lt;/li&gt;
&lt;li&gt;The host sends the link, by chat, e-mail or any other way.&lt;/li&gt;
&lt;li&gt;The guest opens the link and gets a temporary account too.&lt;/li&gt;
&lt;li&gt;The guest presses Enter and joins the same group.&lt;/li&gt;
&lt;li&gt;Realtime tells the host that someone is here.&lt;/li&gt;
&lt;li&gt;The host calls; the guest's browser answers by itself.&lt;/li&gt;
&lt;li&gt;The browsers are connected. From here on the video, the voice, a shared screen and the data channels for the chat, the drawing board and the files travel between the two browsers only.&lt;/li&gt;
&lt;/ol&gt;

&lt;p&gt;Now the code behind the steps. Signing in is a few lines at the top of &lt;code&gt;app.js&lt;/code&gt;. The project ID is the one constant, and &lt;code&gt;autoLogin&lt;/code&gt; brings back the temporary account of an earlier visit, so a reload within half an hour keeps the same account:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight javascript"&gt;&lt;code&gt;&lt;span class="kd"&gt;const&lt;/span&gt; &lt;span class="nx"&gt;PROJECT_ID&lt;/span&gt; &lt;span class="o"&gt;=&lt;/span&gt; &lt;span class="dl"&gt;"&lt;/span&gt;&lt;span class="s2"&gt;5VqfG6NuS163eFEcYy-7Zd6TgdLUaFxseKMT1qveY&lt;/span&gt;&lt;span class="dl"&gt;"&lt;/span&gt;&lt;span class="p"&gt;;&lt;/span&gt;

&lt;span class="kd"&gt;const&lt;/span&gt; &lt;span class="nx"&gt;skapi&lt;/span&gt; &lt;span class="o"&gt;=&lt;/span&gt; &lt;span class="k"&gt;new&lt;/span&gt; &lt;span class="nc"&gt;Skapi&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="nx"&gt;PROJECT_ID&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="p"&gt;{&lt;/span&gt; &lt;span class="na"&gt;autoLogin&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="kc"&gt;true&lt;/span&gt; &lt;span class="p"&gt;});&lt;/span&gt;

&lt;span class="k"&gt;async&lt;/span&gt; &lt;span class="kd"&gt;function&lt;/span&gt; &lt;span class="nf"&gt;signIn&lt;/span&gt;&lt;span class="p"&gt;()&lt;/span&gt; &lt;span class="p"&gt;{&lt;/span&gt;
    &lt;span class="kd"&gt;let&lt;/span&gt; &lt;span class="nx"&gt;user&lt;/span&gt; &lt;span class="o"&gt;=&lt;/span&gt; &lt;span class="k"&gt;await&lt;/span&gt; &lt;span class="nx"&gt;skapi&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;getProfile&lt;/span&gt;&lt;span class="p"&gt;();&lt;/span&gt;
    &lt;span class="k"&gt;if &lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="nx"&gt;user&lt;/span&gt; &lt;span class="o"&gt;&amp;amp;&amp;amp;&lt;/span&gt; &lt;span class="o"&gt;!&lt;/span&gt;&lt;span class="nx"&gt;user&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nx"&gt;anonymous&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt; &lt;span class="p"&gt;{&lt;/span&gt;
        &lt;span class="k"&gt;await&lt;/span&gt; &lt;span class="nx"&gt;skapi&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;logout&lt;/span&gt;&lt;span class="p"&gt;();&lt;/span&gt; &lt;span class="c1"&gt;// a regular user of this origin is logged out here first&lt;/span&gt;
        &lt;span class="nx"&gt;user&lt;/span&gt; &lt;span class="o"&gt;=&lt;/span&gt; &lt;span class="kc"&gt;null&lt;/span&gt;&lt;span class="p"&gt;;&lt;/span&gt;
    &lt;span class="p"&gt;}&lt;/span&gt;
    &lt;span class="k"&gt;if &lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="o"&gt;!&lt;/span&gt;&lt;span class="nx"&gt;user&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt; &lt;span class="nx"&gt;user&lt;/span&gt; &lt;span class="o"&gt;=&lt;/span&gt; &lt;span class="k"&gt;await&lt;/span&gt; &lt;span class="nx"&gt;skapi&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;anonymousLogin&lt;/span&gt;&lt;span class="p"&gt;();&lt;/span&gt;
    &lt;span class="k"&gt;return&lt;/span&gt; &lt;span class="nx"&gt;user&lt;/span&gt;&lt;span class="p"&gt;;&lt;/span&gt;
&lt;span class="p"&gt;}&lt;/span&gt;
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;The room is a realtime group. Without &lt;code&gt;?room=&lt;/code&gt; in the address the page makes a ten-character room id and writes it into the address; that address is the link. Everyone who opens it joins the group of the same name with &lt;code&gt;joinRealtime()&lt;/code&gt;, and the group does two jobs: it says who is there, through the &lt;code&gt;USER_JOINED&lt;/code&gt; and &lt;code&gt;USER_LEFT&lt;/code&gt; messages, and it carries the WebRTC signaling, the offer, the answer and the candidates the two browsers exchange to find a path to each other. A guest looks into the room first with &lt;code&gt;getRealtimeUsers()&lt;/code&gt;, so the waiting room can say "Someone is waiting for you" before they press Enter.&lt;/p&gt;

&lt;p&gt;When two people are in the room, the one with the smaller user id calls. The call is one method, &lt;a href="https://docs.skapi.com/realtime/webRTC.html" rel="noopener noreferrer"&gt;&lt;code&gt;connectRTC()&lt;/code&gt;&lt;/a&gt;, with the camera and the microphone, two named data channels, and a little metadata the other side can read before it answers:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight javascript"&gt;&lt;code&gt;&lt;span class="kd"&gt;const&lt;/span&gt; &lt;span class="nx"&gt;connector&lt;/span&gt; &lt;span class="o"&gt;=&lt;/span&gt; &lt;span class="k"&gt;await&lt;/span&gt; &lt;span class="nx"&gt;skapi&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;connectRTC&lt;/span&gt;&lt;span class="p"&gt;({&lt;/span&gt;
    &lt;span class="nx"&gt;cid&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt;                     &lt;span class="c1"&gt;// the other person's realtime connection&lt;/span&gt;
    &lt;span class="p"&gt;...&lt;/span&gt;&lt;span class="nx"&gt;options&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt;              &lt;span class="c1"&gt;// the camera and the microphone&lt;/span&gt;
    &lt;span class="na"&gt;channels&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="nx"&gt;CHANNELS&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt;      &lt;span class="c1"&gt;// "chat" and "drawing"&lt;/span&gt;
    &lt;span class="na"&gt;meta&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="p"&gt;{&lt;/span&gt; &lt;span class="na"&gt;kind&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="dl"&gt;"&lt;/span&gt;&lt;span class="s2"&gt;video call&lt;/span&gt;&lt;span class="dl"&gt;"&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="nx"&gt;room&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="na"&gt;name&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="nx"&gt;me&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nx"&gt;name&lt;/span&gt; &lt;span class="p"&gt;},&lt;/span&gt;
&lt;span class="p"&gt;},&lt;/span&gt; &lt;span class="nx"&gt;onRTC&lt;/span&gt;&lt;span class="p"&gt;);&lt;/span&gt;

&lt;span class="kd"&gt;const&lt;/span&gt; &lt;span class="nx"&gt;connection&lt;/span&gt; &lt;span class="o"&gt;=&lt;/span&gt; &lt;span class="k"&gt;await&lt;/span&gt; &lt;span class="nx"&gt;connector&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nx"&gt;connection&lt;/span&gt;&lt;span class="p"&gt;;&lt;/span&gt; &lt;span class="c1"&gt;// resolves when the other side picks up&lt;/span&gt;
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;The other browser gets an &lt;code&gt;rtc:incoming&lt;/code&gt; message and answers it with &lt;code&gt;connectRTC()&lt;/code&gt; of its own. In a normal app that is where you would show an incoming call screen; here the link is the invitation, so the guest's browser accepts at once and nobody clicks anything. From then on &lt;code&gt;onRTC&lt;/code&gt; delivers the media tracks and the data channel messages. The chat sends small JSON messages over the &lt;code&gt;chat&lt;/code&gt; channel, which also carries your name and your mute state. The drawing board sends every stroke over the &lt;code&gt;drawing&lt;/code&gt; channel as fractions of a 1600 by 1000 canvas, so both boards show the same picture whatever the screen size. A file goes through &lt;code&gt;sendFile()&lt;/code&gt; on the same connection: the SDK chunks it over a channel of its own, reports progress on both sides, and hands the other browser a file.&lt;/p&gt;

&lt;p&gt;That is the whole mechanism. There is no media server, no signaling server of your own, no TURN configuration to write, and no backend code. The browser already knows how to do a video call; what it needs is a way to find the other browser and a safe place to exchange the handshake, and that is what a Skapi project gives it.&lt;/p&gt;

&lt;h2&gt;
  
  
  Meet anonymousLogin()
&lt;/h2&gt;

&lt;p&gt;Everything in Skapi that is worth protecting, private records, files, realtime, RTC, needs a logged-in user. That is right for an app with accounts, and wrong for a page you want to open and use in three seconds. &lt;code&gt;anonymousLogin()&lt;/code&gt; is the answer. One call signs the visitor in with a temporary anonymous account: no e-mail, no password, no form.&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight javascript"&gt;&lt;code&gt;&lt;span class="kd"&gt;let&lt;/span&gt; &lt;span class="nx"&gt;user&lt;/span&gt; &lt;span class="o"&gt;=&lt;/span&gt; &lt;span class="k"&gt;await&lt;/span&gt; &lt;span class="nx"&gt;skapi&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;anonymousLogin&lt;/span&gt;&lt;span class="p"&gt;();&lt;/span&gt;
&lt;span class="nx"&gt;console&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;log&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="nx"&gt;user&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nx"&gt;anonymous&lt;/span&gt;&lt;span class="p"&gt;);&lt;/span&gt; &lt;span class="c1"&gt;// true&lt;/span&gt;
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;From that moment the visitor is a logged-in user. &lt;code&gt;skapi.user&lt;/code&gt; is set, the login listeners run, and every method that needs a login works: records, files, realtime, RTC, subscriptions, tickets. The account carries &lt;code&gt;anonymous: true&lt;/code&gt; and access group 1, and it counts as a user of your project while it exists.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fnp5xu2wmpdrlld35mjwf.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fnp5xu2wmpdrlld35mjwf.png" alt="skapi users" width="800" height="247"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;&lt;em&gt;An anonymous account lives while a page holds its session, and goes away on its own after the visitor leaves.&lt;/em&gt;&lt;/p&gt;

&lt;p&gt;The SDK keeps the account alive with a keep-alive every 10 minutes while a page holds its session, and once more when a page load restores the session from storage. The account is deleted 30 minutes after the last keep-alive, so reloading the page or coming back within half an hour keeps the account and its data, a tab in the background keeps it too, and a laptop that was asleep for an hour comes back to a deleted one. In that case the next keep-alive learns it, signs the page out, and the page can call &lt;code&gt;anonymousLogin()&lt;/code&gt; again for a fresh account. &lt;code&gt;logout()&lt;/code&gt; deletes the account at once. When the account goes, everything it owns goes with it: its records, files, subscriptions and realtime state.&lt;/p&gt;

&lt;p&gt;A temporary account stays temporary. It cannot edit its profile, change or reset a password, verify an e-mail or a phone number, subscribe to newsletters, or disable itself. Each of these is refused by the server as well as by the SDK, so there is no way around it.&lt;/p&gt;

&lt;p&gt;Anonymous login is off until you turn on &lt;strong&gt;Anonymous Login&lt;/strong&gt; in your project's settings, and it works whether or not &lt;strong&gt;Allow Signup&lt;/strong&gt; is on, so a project can close signups and still let visitors in. The full description is in the documentation: &lt;a href="https://docs.skapi.com/authentication/anonymous-login.html" rel="noopener noreferrer"&gt;Anonymous Login&lt;/a&gt;.&lt;/p&gt;

&lt;p&gt;In Video Call this is the first thing the page does, before anything else, and it is what makes the rest possible. Two anonymous accounts can join a realtime group, exchange a WebRTC handshake, and open a connection to each other, and when the call is over and the tabs are closed, the two accounts are swept away half an hour later and the project is as clean as it was before. There is nothing to moderate, nothing to delete, and no user list that grows with every visitor.&lt;/p&gt;

&lt;h2&gt;
  
  
  Make it yours
&lt;/h2&gt;

&lt;p&gt;The app is four files: &lt;code&gt;index.html&lt;/code&gt; for the page, &lt;code&gt;style.css&lt;/code&gt; for the look, &lt;code&gt;app.js&lt;/code&gt; for everything it does, and &lt;code&gt;markdown.js&lt;/code&gt; for the chat. There is no build step, so it can be hosted anywhere that serves static files, including the web hosting that comes with your Skapi project. The &lt;code&gt;skapi-js&lt;/code&gt; library is loaded from a CDN. Browsers allow the camera, the microphone and calls only on &lt;code&gt;localhost&lt;/code&gt; or over https, so serve it over https when you put it on a server.&lt;/p&gt;

&lt;p&gt;To run it on your computer you need Node.js. In the app's folder:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight shell"&gt;&lt;code&gt;npm &lt;span class="nb"&gt;install
&lt;/span&gt;npm run dev
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;Then open &lt;code&gt;http://localhost:4000/&lt;/code&gt;. Any other static file server works just as well.&lt;/p&gt;

&lt;p&gt;To run it on a project of your own:&lt;/p&gt;

&lt;ol&gt;
&lt;li&gt;Create a free account at &lt;a href="https://www.skapi.com" rel="noopener noreferrer"&gt;www.skapi.com&lt;/a&gt; and create a project.&lt;/li&gt;
&lt;li&gt;Open the project's &lt;strong&gt;Settings&lt;/strong&gt; and turn &lt;strong&gt;Anonymous Login&lt;/strong&gt; on.&lt;/li&gt;
&lt;li&gt;Copy the project's &lt;strong&gt;Project ID&lt;/strong&gt; from the project page.&lt;/li&gt;
&lt;li&gt;Open &lt;code&gt;app.js&lt;/code&gt; and paste it into the one line near the top: &lt;code&gt;const PROJECT_ID = "your-project-id";&lt;/code&gt;
&lt;/li&gt;
&lt;/ol&gt;

&lt;p&gt;That is the only change needed. Upload the folder to your web host and share the address. The colours, sizes and fonts are variables at the top of &lt;code&gt;style.css&lt;/code&gt;, the icons are drawn inline in &lt;code&gt;index.html&lt;/code&gt;, and each part of &lt;code&gt;app.js&lt;/code&gt; starts with a comment that says what it does, so changing the look or the behaviour is a matter of reading one section. Take the drawing board out, keep only the chat, add a third person, change the rules for who may enter: the file is yours.&lt;/p&gt;

&lt;h2&gt;
  
  
  Where to go next
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;The app: &lt;a href="https://videocall.skapi.com/" rel="noopener noreferrer"&gt;videocall.skapi.com&lt;/a&gt;. Free for anyone, no account, nothing to install.&lt;/li&gt;
&lt;li&gt;The code: &lt;a href="https://github.com/broadwayinc/skapi-video-call" rel="noopener noreferrer"&gt;github.com/broadwayinc/skapi-video-call&lt;/a&gt;. The README is the user manual, with every screen explained.&lt;/li&gt;
&lt;li&gt;The feature: &lt;a href="https://docs.skapi.com/authentication/anonymous-login.html" rel="noopener noreferrer"&gt;Anonymous Login&lt;/a&gt; in the Skapi documentation.&lt;/li&gt;
&lt;li&gt;The methods the app is built on: &lt;a href="https://docs.skapi.com/realtime/webRTC.html" rel="noopener noreferrer"&gt;WebRTC&lt;/a&gt; and &lt;a href="https://docs.skapi.com/realtime/introduction.html" rel="noopener noreferrer"&gt;Realtime&lt;/a&gt;.&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;Open &lt;a href="https://videocall.skapi.com/" rel="noopener noreferrer"&gt;videocall.skapi.com&lt;/a&gt;, send the link, talk. Then open the files and see how little it took.&lt;/p&gt;

</description>
      <category>ai</category>
      <category>webdev</category>
      <category>frontend</category>
      <category>skapi</category>
    </item>
    <item>
      <title>Database built for AI agents</title>
      <dc:creator>Skapi</dc:creator>
      <pubDate>Mon, 26 Oct 2026 07:53:27 +0000</pubDate>
      <link>https://dev.to/skapi_api/database-built-for-ai-agents-o3k</link>
      <guid>https://dev.to/skapi_api/database-built-for-ai-agents-o3k</guid>
      <description>&lt;h2&gt;
  
  
  No schema, no setup, no scans: how the Skapi database works, and why it is built for AI agents
&lt;/h2&gt;

&lt;p&gt;The question we get more than any other is some version of "how do I set up my database?" Where is the schema designer. Where do I define the tables. What about indexes, migrations, the query planner. The answer is that you do not. There is nothing to set up. You write a record, and the table exists. You delete the last record, and the table is gone. If that sounds like something that cannot possibly hold up under real load, this article is for you. It explains the model behind the Skapi database, schema-free and index-on-write: why it does the work you reach for a relational database to do, why it never falls into the full-table scan that quietly drains production budgets, why the same mechanism is what keeps your data secure, and why it is the right shape for a world where AI agents write most of the code.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fcdn.broadwayinc.com%2Fimg%2Fskapi-logo-cut-small.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fcdn.broadwayinc.com%2Fimg%2Fskapi-logo-cut-small.png" alt="Skapi" width="244" height="300"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;h2&gt;
  
  
  The question, answered once
&lt;/h2&gt;

&lt;p&gt;A Skapi project has no database to set up, because there is no moment at which structure is decided ahead of data.&lt;/p&gt;

&lt;p&gt;Tables are not created. A table appears when the first record is written into it and disappears when the last one leaves. Indexes are not defined. An index is an argument you pass when you write a record. Access rules are not configured. They are an argument too. There are no migrations, because there is nothing to migrate: a record written today with a new field does not care what the record next to it looks like, and the database does not care either.&lt;/p&gt;

&lt;p&gt;So when you ask "how should I design my schema", the honest answer is this. Write the records your product needs, and say, on each one, how you want to find it again. That is the whole discipline. Everything below is about why that is enough.&lt;/p&gt;

&lt;h2&gt;
  
  
  Where the confusion comes from
&lt;/h2&gt;

&lt;p&gt;If you come from traditional backend engineering, you have been trained on two ways a database decides what data looks like.&lt;/p&gt;

&lt;p&gt;The first is &lt;strong&gt;schema-on-write&lt;/strong&gt;. You design the structure before any data exists: tables, columns, types, constraints, indexes. The database refuses anything that does not fit. This is the relational model, and it is very good at protecting the integrity of data. What it does not protect is the cost of reading it. Every query is handed to a planner at runtime, and the planner decides how to find the rows. If your condition happens to be covered by an index, it seeks. If it is not, it scans the whole table. Both are legal. Nothing warns you. The scan is the silent default for any question nobody anticipated when the schema was drawn.&lt;/p&gt;

&lt;p&gt;The second is &lt;strong&gt;schema-on-read&lt;/strong&gt;. You store whatever arrives and decide what it means when you query it. This is the data lake, and it is also what a document store becomes when its indexes are an afterthought. There is no structure to maintain, which is pleasant, and every read is a scan by definition, which is not.&lt;/p&gt;

&lt;p&gt;Both models share one trait. The cost of a read is decided at read time, by machinery you do not see, over data you wrote earlier without saying how you would read it. The expensive decision is made at the moment you have the least information and the least control.&lt;/p&gt;

&lt;h2&gt;
  
  
  Index-on-write
&lt;/h2&gt;

&lt;p&gt;Skapi moves that decision to the one moment you know the most about a record: when you write it.&lt;/p&gt;

&lt;p&gt;When you post a record, you say a few things alongside the data. Which table it belongs to. Which access group may read it. Optionally, one index name and value to find it by. Optionally, tags. Optionally, the record it references. The database stores the record under keys built from those declarations, in sorted order, typed so that numbers sort as numbers and strings as strings.&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight javascript"&gt;&lt;code&gt;&lt;span class="k"&gt;await&lt;/span&gt; &lt;span class="nx"&gt;skapi&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;postRecord&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;
    &lt;span class="p"&gt;{&lt;/span&gt; &lt;span class="na"&gt;title&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="s1"&gt;Dukkha&lt;/span&gt;&lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="na"&gt;artist&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="s1"&gt;Asian Spice House&lt;/span&gt;&lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="na"&gt;tracks&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="mi"&gt;7&lt;/span&gt; &lt;span class="p"&gt;},&lt;/span&gt;
    &lt;span class="p"&gt;{&lt;/span&gt;
        &lt;span class="na"&gt;table&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="p"&gt;{&lt;/span&gt; &lt;span class="na"&gt;name&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="s1"&gt;Albums&lt;/span&gt;&lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="na"&gt;access_group&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="s1"&gt;public&lt;/span&gt;&lt;span class="dl"&gt;'&lt;/span&gt; &lt;span class="p"&gt;},&lt;/span&gt;
        &lt;span class="na"&gt;index&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="p"&gt;{&lt;/span&gt; &lt;span class="na"&gt;name&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="s1"&gt;year&lt;/span&gt;&lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="na"&gt;value&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="mi"&gt;2023&lt;/span&gt; &lt;span class="p"&gt;},&lt;/span&gt;
        &lt;span class="na"&gt;tags&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="p"&gt;[&lt;/span&gt;&lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="s1"&gt;Indie&lt;/span&gt;&lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="s1"&gt;Experimental&lt;/span&gt;&lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="p"&gt;]&lt;/span&gt;
    &lt;span class="p"&gt;}&lt;/span&gt;
&lt;span class="p"&gt;);&lt;/span&gt;
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;That is the schema. It is local to this record, it was decided by the code that had the record in hand, and it is already applied by the time the write returns.&lt;/p&gt;

&lt;p&gt;A query is the mirror image: a table, an access group, and at most one index condition or range.&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight javascript"&gt;&lt;code&gt;&lt;span class="k"&gt;await&lt;/span&gt; &lt;span class="nx"&gt;skapi&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;getRecords&lt;/span&gt;&lt;span class="p"&gt;({&lt;/span&gt;
    &lt;span class="na"&gt;table&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="p"&gt;{&lt;/span&gt; &lt;span class="na"&gt;name&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="s1"&gt;Albums&lt;/span&gt;&lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="na"&gt;access_group&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="s1"&gt;public&lt;/span&gt;&lt;span class="dl"&gt;'&lt;/span&gt; &lt;span class="p"&gt;},&lt;/span&gt;
    &lt;span class="na"&gt;index&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="p"&gt;{&lt;/span&gt; &lt;span class="na"&gt;name&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="s1"&gt;year&lt;/span&gt;&lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="na"&gt;value&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="mi"&gt;2010&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="na"&gt;condition&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="s1"&gt;&amp;gt;=&lt;/span&gt;&lt;span class="dl"&gt;'&lt;/span&gt; &lt;span class="p"&gt;}&lt;/span&gt;
&lt;span class="p"&gt;});&lt;/span&gt;
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;Every such query maps to one contiguous range of sorted keys. The read is a seek to the start of the range and a walk to the end of it, a page at a time. Nothing outside the range is touched. There is no planner, because there is nothing to plan.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fflfltu0p868xx81phop8.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fflfltu0p868xx81phop8.png" alt="query maps" width="800" height="320"&gt;&lt;/a&gt;&lt;br&gt;
&lt;em&gt;One record, written once. The table, access group and index you pass become a sorted key, and a query reads one range of those keys. Rows outside the range are never touched.&lt;/em&gt;&lt;/p&gt;

&lt;div class="table-wrapper-paragraph"&gt;&lt;table&gt;
&lt;thead&gt;
&lt;tr&gt;
&lt;th&gt;&lt;/th&gt;
&lt;th&gt;Structure is decided&lt;/th&gt;
&lt;th&gt;Access path is decided&lt;/th&gt;
&lt;th&gt;What a read costs&lt;/th&gt;
&lt;/tr&gt;
&lt;/thead&gt;
&lt;tbody&gt;
&lt;tr&gt;
&lt;td&gt;Schema-on-write (relational)&lt;/td&gt;
&lt;td&gt;Before any data, for the whole table&lt;/td&gt;
&lt;td&gt;At query time, by the planner&lt;/td&gt;
&lt;td&gt;A seek if an index happens to cover it, a scan if not&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Schema-on-read (lakes, loose document stores)&lt;/td&gt;
&lt;td&gt;At query time&lt;/td&gt;
&lt;td&gt;At query time&lt;/td&gt;
&lt;td&gt;A scan&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;&lt;strong&gt;Index-on-write (Skapi)&lt;/strong&gt;&lt;/td&gt;
&lt;td&gt;&lt;strong&gt;Per record, when it is written&lt;/strong&gt;&lt;/td&gt;
&lt;td&gt;&lt;strong&gt;When it is written&lt;/strong&gt;&lt;/td&gt;
&lt;td&gt;&lt;strong&gt;One key range, proportional to the result&lt;/strong&gt;&lt;/td&gt;
&lt;/tr&gt;
&lt;/tbody&gt;
&lt;/table&gt;&lt;/div&gt;
&lt;h2&gt;
  
  
  The expensive query cannot be written
&lt;/h2&gt;

&lt;p&gt;Here is the theory in one line. A scan costs in proportion to the size of the table. A seek costs in proportion to the size of the result. Everything that goes wrong with database bills lives in the gap between those two.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fypfh2a0b8mda7vusngn3.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fypfh2a0b8mda7vusngn3.png" alt="database bills" width="799" height="429"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;&lt;em&gt;A scan reads every row, so its cost grows with the table. A seek reads the matching range, so its cost grows with the result. On Skapi, only the seek exists.&lt;/em&gt;&lt;/p&gt;

&lt;p&gt;Every experienced engineer has seen the pattern. A query runs fine in development against ten thousand rows. It ships. The table grows to fifty million rows, and the same query is now five thousand times more expensive, because it was always a scan and nobody noticed while the table was small. On a database billed per read, that is not a slow page. That is an invoice. The fix is an index that someone had to think of in advance, which is exactly the thinking the schema was supposed to force, and did not.&lt;/p&gt;

&lt;p&gt;On Skapi the scan is not slow. It does not exist. There is no way to ask the database for "every record whose data contains some value", because the data is not what the database reads. The keys are. If you want to find records by a value, you index on that value when you write. If you did not, the query is not expensive, it is not a query.&lt;/p&gt;

&lt;p&gt;The consequence is that the cost of any read is proportional to what it returns, not to how much you have stored. A table of ten records and a table of ten million answer the same query at the same cost per record returned. There is no slow-query log to read, no planner to second-guess, and no index advisor telling you what you should have done last month. A junior developer and a senior one write the same query, because there is only one kind of query to write.&lt;/p&gt;
&lt;h2&gt;
  
  
  Everything you build a relational database for
&lt;/h2&gt;

&lt;p&gt;The objection at this point is that a single index per record cannot possibly cover what a relational database does. It covers what you actually build. Every item below is a key read, and each one is a few lines of skapi-js.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Find by id.&lt;/strong&gt; Every record has an id, and you can give it your own: an order number, a slug, a SKU. Both are direct lookups.&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight javascript"&gt;&lt;code&gt;&lt;span class="k"&gt;await&lt;/span&gt; &lt;span class="nx"&gt;skapi&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;postRecord&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="nx"&gt;order&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="p"&gt;{&lt;/span&gt;
    &lt;span class="na"&gt;table&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="p"&gt;{&lt;/span&gt; &lt;span class="na"&gt;name&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="s1"&gt;Orders&lt;/span&gt;&lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="na"&gt;access_group&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="s1"&gt;authorized&lt;/span&gt;&lt;span class="dl"&gt;'&lt;/span&gt; &lt;span class="p"&gt;},&lt;/span&gt;
    &lt;span class="na"&gt;unique_id&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="s1"&gt;order-1042&lt;/span&gt;&lt;span class="dl"&gt;'&lt;/span&gt;
&lt;span class="p"&gt;});&lt;/span&gt;

&lt;span class="k"&gt;await&lt;/span&gt; &lt;span class="nx"&gt;skapi&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;getRecords&lt;/span&gt;&lt;span class="p"&gt;({&lt;/span&gt; &lt;span class="na"&gt;unique_id&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="s1"&gt;order-1042&lt;/span&gt;&lt;span class="dl"&gt;'&lt;/span&gt; &lt;span class="p"&gt;});&lt;/span&gt;
&lt;span class="k"&gt;await&lt;/span&gt; &lt;span class="nx"&gt;skapi&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;getRecords&lt;/span&gt;&lt;span class="p"&gt;({&lt;/span&gt; &lt;span class="na"&gt;record_id&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="nx"&gt;order&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nx"&gt;record_id&lt;/span&gt; &lt;span class="p"&gt;});&lt;/span&gt;
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;&lt;strong&gt;Newest first, oldest first.&lt;/strong&gt; Every record is ordered by when it was uploaded and when it was last updated. Both are reserved index names, so "everything from the last seven days, newest first" is a range read.&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight javascript"&gt;&lt;code&gt;&lt;span class="kd"&gt;const&lt;/span&gt; &lt;span class="nx"&gt;since&lt;/span&gt; &lt;span class="o"&gt;=&lt;/span&gt; &lt;span class="nb"&gt;Date&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;now&lt;/span&gt;&lt;span class="p"&gt;()&lt;/span&gt; &lt;span class="o"&gt;-&lt;/span&gt; &lt;span class="mi"&gt;7&lt;/span&gt; &lt;span class="o"&gt;*&lt;/span&gt; &lt;span class="mi"&gt;24&lt;/span&gt; &lt;span class="o"&gt;*&lt;/span&gt; &lt;span class="mi"&gt;60&lt;/span&gt; &lt;span class="o"&gt;*&lt;/span&gt; &lt;span class="mi"&gt;60&lt;/span&gt; &lt;span class="o"&gt;*&lt;/span&gt; &lt;span class="mi"&gt;1000&lt;/span&gt;&lt;span class="p"&gt;;&lt;/span&gt;

&lt;span class="k"&gt;await&lt;/span&gt; &lt;span class="nx"&gt;skapi&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;getRecords&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;
    &lt;span class="p"&gt;{&lt;/span&gt;
        &lt;span class="na"&gt;table&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="p"&gt;{&lt;/span&gt; &lt;span class="na"&gt;name&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="s1"&gt;Orders&lt;/span&gt;&lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="na"&gt;access_group&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="s1"&gt;authorized&lt;/span&gt;&lt;span class="dl"&gt;'&lt;/span&gt; &lt;span class="p"&gt;},&lt;/span&gt;
        &lt;span class="na"&gt;index&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="p"&gt;{&lt;/span&gt; &lt;span class="na"&gt;name&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="s1"&gt;$uploaded&lt;/span&gt;&lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="na"&gt;value&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="nx"&gt;since&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="na"&gt;condition&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="s1"&gt;&amp;gt;=&lt;/span&gt;&lt;span class="dl"&gt;'&lt;/span&gt; &lt;span class="p"&gt;}&lt;/span&gt;
    &lt;span class="p"&gt;},&lt;/span&gt;
    &lt;span class="p"&gt;{&lt;/span&gt; &lt;span class="na"&gt;ascending&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="kc"&gt;false&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="na"&gt;limit&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="mi"&gt;50&lt;/span&gt; &lt;span class="p"&gt;}&lt;/span&gt;
&lt;span class="p"&gt;);&lt;/span&gt;
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;&lt;strong&gt;Conditions and ranges.&lt;/strong&gt; Equal, greater, greater or equal, less, less or equal, and a range between two values. On a string, the same operators give you "starts with" and "ends with".&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight javascript"&gt;&lt;code&gt;&lt;span class="c1"&gt;// albums released from 1960 to 1970&lt;/span&gt;
&lt;span class="k"&gt;await&lt;/span&gt; &lt;span class="nx"&gt;skapi&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;getRecords&lt;/span&gt;&lt;span class="p"&gt;({&lt;/span&gt;
    &lt;span class="na"&gt;table&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="p"&gt;{&lt;/span&gt; &lt;span class="na"&gt;name&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="s1"&gt;Albums&lt;/span&gt;&lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="na"&gt;access_group&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="s1"&gt;public&lt;/span&gt;&lt;span class="dl"&gt;'&lt;/span&gt; &lt;span class="p"&gt;},&lt;/span&gt;
    &lt;span class="na"&gt;index&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="p"&gt;{&lt;/span&gt; &lt;span class="na"&gt;name&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="s1"&gt;year&lt;/span&gt;&lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="na"&gt;value&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="mi"&gt;1960&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="na"&gt;range&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="mi"&gt;1970&lt;/span&gt; &lt;span class="p"&gt;}&lt;/span&gt;
&lt;span class="p"&gt;});&lt;/span&gt;

&lt;span class="c1"&gt;// titles that end with "Gilberto" ('&amp;gt;=' on a string is "starts with")&lt;/span&gt;
&lt;span class="k"&gt;await&lt;/span&gt; &lt;span class="nx"&gt;skapi&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;getRecords&lt;/span&gt;&lt;span class="p"&gt;({&lt;/span&gt;
    &lt;span class="na"&gt;table&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="p"&gt;{&lt;/span&gt; &lt;span class="na"&gt;name&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="s1"&gt;Albums&lt;/span&gt;&lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="na"&gt;access_group&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="s1"&gt;public&lt;/span&gt;&lt;span class="dl"&gt;'&lt;/span&gt; &lt;span class="p"&gt;},&lt;/span&gt;
    &lt;span class="na"&gt;index&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="p"&gt;{&lt;/span&gt; &lt;span class="na"&gt;name&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="s1"&gt;title&lt;/span&gt;&lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="na"&gt;value&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="s1"&gt;Gilberto&lt;/span&gt;&lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="na"&gt;condition&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="s1"&gt;&amp;lt;=&lt;/span&gt;&lt;span class="dl"&gt;'&lt;/span&gt; &lt;span class="p"&gt;}&lt;/span&gt;
&lt;span class="p"&gt;});&lt;/span&gt;
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;&lt;strong&gt;Relations.&lt;/strong&gt; A record can reference another record. Comments reference a post, order items reference an order, a like references what was liked. Fetch everything that references a record, and the relationship is already there. A join is the engine discovering a relationship at query time. A reference is the relationship recorded at write time, which is the only moment anyone actually knew it.&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight javascript"&gt;&lt;code&gt;&lt;span class="kd"&gt;const&lt;/span&gt; &lt;span class="nx"&gt;post&lt;/span&gt; &lt;span class="o"&gt;=&lt;/span&gt; &lt;span class="k"&gt;await&lt;/span&gt; &lt;span class="nx"&gt;skapi&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;postRecord&lt;/span&gt;&lt;span class="p"&gt;({&lt;/span&gt; &lt;span class="na"&gt;text&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="s1"&gt;What do you think?&lt;/span&gt;&lt;span class="dl"&gt;'&lt;/span&gt; &lt;span class="p"&gt;},&lt;/span&gt; &lt;span class="p"&gt;{&lt;/span&gt;
    &lt;span class="na"&gt;table&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="p"&gt;{&lt;/span&gt; &lt;span class="na"&gt;name&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="s1"&gt;Posts&lt;/span&gt;&lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="na"&gt;access_group&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="s1"&gt;public&lt;/span&gt;&lt;span class="dl"&gt;'&lt;/span&gt; &lt;span class="p"&gt;}&lt;/span&gt;
&lt;span class="p"&gt;});&lt;/span&gt;

&lt;span class="k"&gt;await&lt;/span&gt; &lt;span class="nx"&gt;skapi&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;postRecord&lt;/span&gt;&lt;span class="p"&gt;({&lt;/span&gt; &lt;span class="na"&gt;text&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="s1"&gt;I like it!&lt;/span&gt;&lt;span class="dl"&gt;'&lt;/span&gt; &lt;span class="p"&gt;},&lt;/span&gt; &lt;span class="p"&gt;{&lt;/span&gt;
    &lt;span class="na"&gt;table&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="p"&gt;{&lt;/span&gt; &lt;span class="na"&gt;name&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="s1"&gt;Comments&lt;/span&gt;&lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="na"&gt;access_group&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="s1"&gt;public&lt;/span&gt;&lt;span class="dl"&gt;'&lt;/span&gt; &lt;span class="p"&gt;},&lt;/span&gt;
    &lt;span class="na"&gt;reference&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="nx"&gt;post&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nx"&gt;record_id&lt;/span&gt;
&lt;span class="p"&gt;});&lt;/span&gt;

&lt;span class="c1"&gt;// every comment on that post&lt;/span&gt;
&lt;span class="k"&gt;await&lt;/span&gt; &lt;span class="nx"&gt;skapi&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;getRecords&lt;/span&gt;&lt;span class="p"&gt;({&lt;/span&gt;
    &lt;span class="na"&gt;table&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="p"&gt;{&lt;/span&gt; &lt;span class="na"&gt;name&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="s1"&gt;Comments&lt;/span&gt;&lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="na"&gt;access_group&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="s1"&gt;public&lt;/span&gt;&lt;span class="dl"&gt;'&lt;/span&gt; &lt;span class="p"&gt;},&lt;/span&gt;
    &lt;span class="na"&gt;reference&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="nx"&gt;post&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nx"&gt;record_id&lt;/span&gt;
&lt;span class="p"&gt;});&lt;/span&gt;
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fi4fzzfenruichjuaevt0.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fi4fzzfenruichjuaevt0.png" alt="post is written once" width="800" height="292"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;&lt;em&gt;A post is written once. Each comment and like points at it, and the post carries a count of how many records point at it. Fetching by reference reads those records and nothing else.&lt;/em&gt;&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Counting relations.&lt;/strong&gt; How many records reference a record is itself an index. "The most commented posts" is a range read, not a GROUP BY over a comments table.&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight javascript"&gt;&lt;code&gt;&lt;span class="c1"&gt;// posts with more than ten comments, most commented first&lt;/span&gt;
&lt;span class="k"&gt;await&lt;/span&gt; &lt;span class="nx"&gt;skapi&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;getRecords&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;
    &lt;span class="p"&gt;{&lt;/span&gt;
        &lt;span class="na"&gt;table&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="p"&gt;{&lt;/span&gt; &lt;span class="na"&gt;name&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="s1"&gt;Posts&lt;/span&gt;&lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="na"&gt;access_group&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="s1"&gt;public&lt;/span&gt;&lt;span class="dl"&gt;'&lt;/span&gt; &lt;span class="p"&gt;},&lt;/span&gt;
        &lt;span class="na"&gt;index&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="p"&gt;{&lt;/span&gt; &lt;span class="na"&gt;name&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="s1"&gt;$referenced_count&lt;/span&gt;&lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="na"&gt;value&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="mi"&gt;10&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="na"&gt;condition&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="s1"&gt;&amp;gt;&lt;/span&gt;&lt;span class="dl"&gt;'&lt;/span&gt; &lt;span class="p"&gt;}&lt;/span&gt;
    &lt;span class="p"&gt;},&lt;/span&gt;
    &lt;span class="p"&gt;{&lt;/span&gt; &lt;span class="na"&gt;ascending&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="kc"&gt;false&lt;/span&gt; &lt;span class="p"&gt;}&lt;/span&gt;
&lt;span class="p"&gt;);&lt;/span&gt;
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;&lt;strong&gt;Hierarchy.&lt;/strong&gt; A compound index name such as &lt;code&gt;Band.AsianSpiceHouse.year&lt;/code&gt; can be queried at any level: every band, bands by name, or one band's albums by year. Here is the write, and the index part of three queries on the same table.&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight javascript"&gt;&lt;code&gt;&lt;span class="k"&gt;await&lt;/span&gt; &lt;span class="nx"&gt;skapi&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;postRecord&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="nx"&gt;album&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="p"&gt;{&lt;/span&gt;
    &lt;span class="na"&gt;table&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="p"&gt;{&lt;/span&gt; &lt;span class="na"&gt;name&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="s1"&gt;Albums&lt;/span&gt;&lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="na"&gt;access_group&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="s1"&gt;public&lt;/span&gt;&lt;span class="dl"&gt;'&lt;/span&gt; &lt;span class="p"&gt;},&lt;/span&gt;
    &lt;span class="na"&gt;index&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="p"&gt;{&lt;/span&gt; &lt;span class="na"&gt;name&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="s1"&gt;Band.AsianSpiceHouse.year&lt;/span&gt;&lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="na"&gt;value&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="mi"&gt;2023&lt;/span&gt; &lt;span class="p"&gt;}&lt;/span&gt;
&lt;span class="p"&gt;});&lt;/span&gt;

&lt;span class="c1"&gt;// every album by a band&lt;/span&gt;
&lt;span class="nl"&gt;index&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="p"&gt;{&lt;/span&gt; &lt;span class="na"&gt;name&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="s1"&gt;Band.&lt;/span&gt;&lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="na"&gt;value&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="dl"&gt;''&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="na"&gt;condition&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="s1"&gt;&amp;gt;&lt;/span&gt;&lt;span class="dl"&gt;'&lt;/span&gt; &lt;span class="p"&gt;}&lt;/span&gt;

&lt;span class="c1"&gt;// bands whose name starts with "Asian"&lt;/span&gt;
&lt;span class="nl"&gt;index&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="p"&gt;{&lt;/span&gt; &lt;span class="na"&gt;name&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="s1"&gt;Band.&lt;/span&gt;&lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="na"&gt;value&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="s1"&gt;Asian&lt;/span&gt;&lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="na"&gt;condition&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="s1"&gt;&amp;gt;=&lt;/span&gt;&lt;span class="dl"&gt;'&lt;/span&gt; &lt;span class="p"&gt;}&lt;/span&gt;

&lt;span class="c1"&gt;// one band's albums released after 2010&lt;/span&gt;
&lt;span class="nl"&gt;index&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="p"&gt;{&lt;/span&gt; &lt;span class="na"&gt;name&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="s1"&gt;Band.AsianSpiceHouse.year&lt;/span&gt;&lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="na"&gt;value&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="mi"&gt;2010&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="na"&gt;condition&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="s1"&gt;&amp;gt;&lt;/span&gt;&lt;span class="dl"&gt;'&lt;/span&gt; &lt;span class="p"&gt;}&lt;/span&gt;
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;&lt;strong&gt;Tags.&lt;/strong&gt; Labels that cut across indexes. A tag narrows any query, and every tag carries its own record count.&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight javascript"&gt;&lt;code&gt;&lt;span class="k"&gt;await&lt;/span&gt; &lt;span class="nx"&gt;skapi&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;postRecord&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="nx"&gt;album&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="p"&gt;{&lt;/span&gt;
    &lt;span class="na"&gt;table&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="p"&gt;{&lt;/span&gt; &lt;span class="na"&gt;name&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="s1"&gt;Albums&lt;/span&gt;&lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="na"&gt;access_group&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="s1"&gt;public&lt;/span&gt;&lt;span class="dl"&gt;'&lt;/span&gt; &lt;span class="p"&gt;},&lt;/span&gt;
    &lt;span class="na"&gt;index&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="p"&gt;{&lt;/span&gt; &lt;span class="na"&gt;name&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="s1"&gt;year&lt;/span&gt;&lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="na"&gt;value&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="mi"&gt;2023&lt;/span&gt; &lt;span class="p"&gt;},&lt;/span&gt;
    &lt;span class="na"&gt;tags&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="p"&gt;[&lt;/span&gt;&lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="s1"&gt;Indie&lt;/span&gt;&lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="s1"&gt;Experimental&lt;/span&gt;&lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="p"&gt;]&lt;/span&gt;
&lt;span class="p"&gt;});&lt;/span&gt;

&lt;span class="c1"&gt;// albums released after 2020 that are tagged Experimental&lt;/span&gt;
&lt;span class="k"&gt;await&lt;/span&gt; &lt;span class="nx"&gt;skapi&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;getRecords&lt;/span&gt;&lt;span class="p"&gt;({&lt;/span&gt;
    &lt;span class="na"&gt;table&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="p"&gt;{&lt;/span&gt; &lt;span class="na"&gt;name&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="s1"&gt;Albums&lt;/span&gt;&lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="na"&gt;access_group&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="s1"&gt;public&lt;/span&gt;&lt;span class="dl"&gt;'&lt;/span&gt; &lt;span class="p"&gt;},&lt;/span&gt;
    &lt;span class="na"&gt;index&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="p"&gt;{&lt;/span&gt; &lt;span class="na"&gt;name&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="s1"&gt;year&lt;/span&gt;&lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="na"&gt;value&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="mi"&gt;2020&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="na"&gt;condition&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="s1"&gt;&amp;gt;&lt;/span&gt;&lt;span class="dl"&gt;'&lt;/span&gt; &lt;span class="p"&gt;},&lt;/span&gt;
    &lt;span class="na"&gt;tag&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="s1"&gt;Experimental&lt;/span&gt;&lt;span class="dl"&gt;'&lt;/span&gt;
&lt;span class="p"&gt;});&lt;/span&gt;
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;&lt;strong&gt;Aggregates without scanning.&lt;/strong&gt; Record counts per table, per tag and per index. Totals and averages of numeric indexes. The rate of true values for boolean indexes. All of it is kept up to date as records are written, so a rating board does not average votes at read time. The average is already there.&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight javascript"&gt;&lt;code&gt;&lt;span class="c1"&gt;// each vote is a record with a boolean index&lt;/span&gt;
&lt;span class="k"&gt;await&lt;/span&gt; &lt;span class="nx"&gt;skapi&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;postRecord&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="kc"&gt;null&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="p"&gt;{&lt;/span&gt;
    &lt;span class="na"&gt;table&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="p"&gt;{&lt;/span&gt; &lt;span class="na"&gt;name&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="s1"&gt;VoteBoard&lt;/span&gt;&lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="na"&gt;access_group&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="s1"&gt;public&lt;/span&gt;&lt;span class="dl"&gt;'&lt;/span&gt; &lt;span class="p"&gt;},&lt;/span&gt;
    &lt;span class="na"&gt;index&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="p"&gt;{&lt;/span&gt; &lt;span class="na"&gt;name&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="s1"&gt;Vote.Beer&lt;/span&gt;&lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="na"&gt;value&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="kc"&gt;true&lt;/span&gt; &lt;span class="p"&gt;}&lt;/span&gt;
&lt;span class="p"&gt;});&lt;/span&gt;

&lt;span class="k"&gt;await&lt;/span&gt; &lt;span class="nx"&gt;skapi&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;getIndexes&lt;/span&gt;&lt;span class="p"&gt;({&lt;/span&gt; &lt;span class="na"&gt;table&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="s1"&gt;VoteBoard&lt;/span&gt;&lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="na"&gt;index&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="s1"&gt;Vote.Beer&lt;/span&gt;&lt;span class="dl"&gt;'&lt;/span&gt; &lt;span class="p"&gt;});&lt;/span&gt;
&lt;span class="c1"&gt;// average_bool: the share of true votes, boolean_count: how many voted&lt;/span&gt;

&lt;span class="k"&gt;await&lt;/span&gt; &lt;span class="nx"&gt;skapi&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;getTables&lt;/span&gt;&lt;span class="p"&gt;({&lt;/span&gt; &lt;span class="na"&gt;table&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="s1"&gt;Albums&lt;/span&gt;&lt;span class="dl"&gt;'&lt;/span&gt; &lt;span class="p"&gt;});&lt;/span&gt;
&lt;span class="c1"&gt;// number_of_records and size, per table&lt;/span&gt;

&lt;span class="k"&gt;await&lt;/span&gt; &lt;span class="nx"&gt;skapi&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;getTags&lt;/span&gt;&lt;span class="p"&gt;({&lt;/span&gt; &lt;span class="na"&gt;table&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="s1"&gt;Albums&lt;/span&gt;&lt;span class="dl"&gt;'&lt;/span&gt; &lt;span class="p"&gt;});&lt;/span&gt;
&lt;span class="c1"&gt;// every tag in the table with its number_of_records&lt;/span&gt;
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;&lt;strong&gt;Per user.&lt;/strong&gt; Every user's records, per table, and subscriptions and feeds on top of that. A user follows another user, and reads one feed of everything the people they follow have posted to it.&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight javascript"&gt;&lt;code&gt;&lt;span class="c1"&gt;// everything one user posted to a table&lt;/span&gt;
&lt;span class="k"&gt;await&lt;/span&gt; &lt;span class="nx"&gt;skapi&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;getRecords&lt;/span&gt;&lt;span class="p"&gt;({&lt;/span&gt;
    &lt;span class="na"&gt;table&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="p"&gt;{&lt;/span&gt; &lt;span class="na"&gt;name&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="s1"&gt;Posts&lt;/span&gt;&lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="na"&gt;access_group&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="s1"&gt;public&lt;/span&gt;&lt;span class="dl"&gt;'&lt;/span&gt; &lt;span class="p"&gt;},&lt;/span&gt;
    &lt;span class="na"&gt;index&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="p"&gt;{&lt;/span&gt; &lt;span class="na"&gt;name&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="s1"&gt;$user_id&lt;/span&gt;&lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="na"&gt;value&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="nx"&gt;user_id&lt;/span&gt; &lt;span class="p"&gt;}&lt;/span&gt;
&lt;span class="p"&gt;});&lt;/span&gt;

&lt;span class="c1"&gt;// follow a user, then read the feed of everyone you follow&lt;/span&gt;
&lt;span class="k"&gt;await&lt;/span&gt; &lt;span class="nx"&gt;skapi&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;subscribe&lt;/span&gt;&lt;span class="p"&gt;({&lt;/span&gt; &lt;span class="nx"&gt;user_id&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="na"&gt;get_feed&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="kc"&gt;true&lt;/span&gt; &lt;span class="p"&gt;});&lt;/span&gt;
&lt;span class="k"&gt;await&lt;/span&gt; &lt;span class="nx"&gt;skapi&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;getFeed&lt;/span&gt;&lt;span class="p"&gt;({&lt;/span&gt; &lt;span class="na"&gt;access_group&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="s1"&gt;authorized&lt;/span&gt;&lt;span class="dl"&gt;'&lt;/span&gt; &lt;span class="p"&gt;});&lt;/span&gt;
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;Notice what is not on the list: anything that would require reading the whole table to answer. That is not an omission. That is the point.&lt;/p&gt;

&lt;h2&gt;
  
  
  Permission is an address, not a filter
&lt;/h2&gt;

&lt;p&gt;On most backends, access control is a filter. A policy is a predicate added to each query and evaluated per row by the same engine that may or may not have an index for it. It is a step in the read, and like any step it can be slow, and it can be skipped by any path that writes to the database without going through it.&lt;/p&gt;

&lt;p&gt;On Skapi, the access group is part of the key a record is stored under. Public records, authorized records, numbered groups from 0 to 99, admin, private: each is a different address. A query names the group it wants to read. A user in access group 5 can ask for group 5 or any group below it, and the read lands on that group's keys alone. A user in group 3 who asks for group 5 is refused before a single key is read. The rows for group 50 are not filtered out of anyone's result. They are never read, because they do not live where the query is looking. Private records are reachable only by their uploader and by users the uploader explicitly grants, and that grant is an address too.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fha0dzdkdpfygkha2tmhc.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fha0dzdkdpfygkha2tmhc.png" alt="grant is an address" width="800" height="310"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;&lt;em&gt;A query names the access group it wants. The read lands on that group's address and nowhere else. Higher groups and private records are not filtered out of the result. They are never read.&lt;/em&gt;&lt;/p&gt;

&lt;p&gt;This is why security and cost come from the same mechanism. A permission check is not a stage that a badly written query can bypass or a busy engine can make expensive. It is where the data is.&lt;/p&gt;

&lt;h2&gt;
  
  
  Sealed in the browser
&lt;/h2&gt;

&lt;p&gt;The model goes one step further for the data a user keeps to themselves. Turn on record encryption, and the data of a private record is encrypted in the browser before it leaves the device, with keys derived from the user's own password that never leave that device. A provider who dumps the database, the storage bucket and every request log cannot recover the contents without guessing the user's password.&lt;/p&gt;

&lt;p&gt;Index-on-write is what makes this possible without breaking the database. The database reads structure: table, access group, index, tags, reference. It stores content. Because those are separate, the structure can stay readable, so every query above keeps working, while the content is sealed. The same is true of attached files: their bytes are encrypted, their names and sizes are not. The details, and the exact security claim, are in the &lt;a href="https://docs.skapi.com/database/encryption.html" rel="noopener noreferrer"&gt;encryption guide&lt;/a&gt;.&lt;/p&gt;

&lt;h2&gt;
  
  
  Why this is the database for AI agents
&lt;/h2&gt;

&lt;p&gt;Agents write code fast and in volume, and they write it against a backend you did not fully review. A database has to be safe under that kind of writer.&lt;/p&gt;

&lt;p&gt;Against a schema-on-write database, every change an agent makes is global. One more field means a migration, a policy per operation, regenerated types and a redeploy, and the loop starts again on the next request. We &lt;a href="https://www.skapi.com/articles/2026-10-06-00" rel="noopener noreferrer"&gt;measured that loop&lt;/a&gt; in tokens. The deeper reason it costs so much is that a schema is shared state: an agent must hold the whole of it in context to change any of it safely.&lt;/p&gt;

&lt;p&gt;Index-on-write has no shared state. Every decision is local to one write: this record, this table, this index, this access group. The agent decides at the point of maximum information and never has to reconcile the decision with anything else. When it needs to know the shape of the data, it asks the database, which lists tables, indexes and tags with their counts, maintained as the data was written. The schema is discovered from the data instead of being maintained beside it.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fx9e4jor6btc6xln6ymh5.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fx9e4jor6btc6xln6ymh5.png" alt="schema on-write" width="800" height="284"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;&lt;em&gt;Three agents changing a schema-on-write backend all pass through one shared schema and the migrate, policies, types, deploy loop. On index-on-write, each agent writes its own record with its own rules, and nothing has to be reconciled.&lt;/em&gt;&lt;/p&gt;

&lt;p&gt;And the cost guard holds for agents exactly as it holds for people. Agents will write queries you never read. On a relational database, one of them will be a scan. On Skapi it cannot be, because the only query that exists is a key read. The worst an agent can do is fetch a page it did not need.&lt;/p&gt;

&lt;p&gt;For fifty years, database design has meant deciding structure before data and paying for every question nobody anticipated. Agents invert the workflow. Data first, structure as a property of each record, access as part of the address, and a database that stays cheap and safe no matter who, or what, is writing to it. That is index-on-write, and it is what the Skapi database has been from the beginning.&lt;/p&gt;

&lt;h2&gt;
  
  
  Start with a record
&lt;/h2&gt;

&lt;p&gt;Create a Free project, write one record from the browser, and you have a database. The &lt;a href="https://docs.skapi.com/database/create.html" rel="noopener noreferrer"&gt;database guide&lt;/a&gt; shows every argument from the example above, and there is nothing to set up before you begin.&lt;/p&gt;

</description>
      <category>ai</category>
      <category>webdev</category>
      <category>database</category>
      <category>backend</category>
    </item>
    <item>
      <title>Newsletters on Skapi: simpler than Mailchimp</title>
      <dc:creator>Skapi</dc:creator>
      <pubDate>Mon, 19 Oct 2026 07:46:39 +0000</pubDate>
      <link>https://dev.to/skapi_api/newsletters-on-skapi-simpler-than-mailchimp-4ieb</link>
      <guid>https://dev.to/skapi_api/newsletters-on-skapi-simpler-than-mailchimp-4ieb</guid>
      <description>&lt;h2&gt;
  
  
  Newsletters on Skapi: simpler than Mailchimp, for a fraction of the price
&lt;/h2&gt;

&lt;p&gt;A marketing newsletter needs three things: a list, a form that fills it, and a way to send. Skapi ships all three inside the backend you already pay for, and the sending part is the mail app you already have. Here is how it works, click by click, and what it costs next to Mailchimp.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fcdn.broadwayinc.com%2Fimg%2Fskapi-logo-cut-small.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fcdn.broadwayinc.com%2Fimg%2Fskapi-logo-cut-small.png" alt="Skapi" width="244" height="300"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;h2&gt;
  
  
  One page, one address
&lt;/h2&gt;

&lt;p&gt;Open your project and choose &lt;strong&gt;Newsletters&lt;/strong&gt; in the sidebar. The page has a tab per list: &lt;strong&gt;Newsletter&lt;/strong&gt; is the public list anyone can join with an email address, &lt;strong&gt;Service Email&lt;/strong&gt; is the list of your signed-in users, and every named group you create gets a tab of its own.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto/https%3A%2F%2Fcdn.broadwayinc.com%2Farticles%2Fimg%2Fnewsletter-page.webp" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto/https%3A%2F%2Fcdn.broadwayinc.com%2Farticles%2Fimg%2Fnewsletter-page.webp" alt="The Newsletters page of a Skapi project: tabs per list, the sending address, monthly sends, subscribers, and the table of sent newsletters" width="800" height="538"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;&lt;em&gt;The Newsletters page. Every list has its own sending address, its own subscriber count and its own history of sent mail.&lt;/em&gt;&lt;/p&gt;

&lt;p&gt;The first row is the whole sending story. Write your newsletter in your own mail app, the one you use every day, and send it to the address shown there. &lt;strong&gt;[Send]&lt;/strong&gt; opens a new message to that address for you. Skapi forwards the email to every subscriber of that list, adds an unsubscribe link at the bottom, and the table below fills in with the subject, the date, and how many people read it, complained, or bounced.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto/https%3A%2F%2Fcdn.broadwayinc.com%2Farticles%2Fimg%2Fnewsletter-compose.webp" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto/https%3A%2F%2Fcdn.broadwayinc.com%2Farticles%2Fimg%2Fnewsletter-compose.webp" alt="A mail compose window: From the project owner's address, To the newsletter sending address, a subject reading Spring menu is here, a short newsletter with a list of items, a link and an attached image, and a Send button" width="800" height="743"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;&lt;em&gt;Composing a newsletter in a mail app. The address in To is the one the Newsletters page shows, and the email goes out as written, with an unsubscribe link added at the bottom.&lt;/em&gt;&lt;/p&gt;

&lt;p&gt;There is no editor to learn, no template gallery, no campaign builder. If you can write an email, you can send a newsletter.&lt;/p&gt;

&lt;p&gt;Two guardrails sit behind that address. Mail is only accepted from the email address your project is registered with, so knowing the address is not enough to send as you. And the sending address has to pass the sender trust checks (SPF, DKIM and DMARC), which the large email providers already do. Anything else is refused, and Skapi replies to tell you why.&lt;/p&gt;

&lt;p&gt;Before any of this works, register an &lt;strong&gt;email alias&lt;/strong&gt; on your project's Settings page. It becomes the address your newsletters and automated emails are sent from, and the Newsletters page stays closed until you have one.&lt;/p&gt;

&lt;h2&gt;
  
  
  Adding subscribers by hand
&lt;/h2&gt;

&lt;p&gt;Not every subscriber comes through a form. You may be moving a list from another tool, or you may have customers who wrote their address on a sheet at the counter. On the &lt;strong&gt;Subscribers&lt;/strong&gt; row, click &lt;strong&gt;[Add]&lt;/strong&gt;.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto/https%3A%2F%2Fcdn.broadwayinc.com%2Farticles%2Fimg%2Fnewsletter-add.webp" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto/https%3A%2F%2Fcdn.broadwayinc.com%2Farticles%2Fimg%2Fnewsletter-add.webp" alt="The Add Subscribers dialog with three pasted addresses staged and a button reading Add 3 subscribers" width="800" height="538"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;&lt;em&gt;Paste addresses, one per line or separated by commas, check the list, and add them.&lt;/em&gt;&lt;/p&gt;

&lt;p&gt;Paste the addresses, one per line or separated by commas, and click &lt;strong&gt;Add to list&lt;/strong&gt;. Anything malformed is set aside and shown to you, duplicates are folded together, and the addresses that remain are listed so you can remove one before committing. Then &lt;strong&gt;Add 3 subscribers&lt;/strong&gt;, or however many you pasted.&lt;/p&gt;

&lt;p&gt;Addresses added this way are subscribed right away, without a confirmation mail, so use it for people who have already asked to hear from you. Everyone else should come in through the form below, which confirms the address first.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;[View]&lt;/strong&gt; opens the list itself, with a search that matches the start of an address.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto/https%3A%2F%2Fcdn.broadwayinc.com%2Farticles%2Fimg%2Fnewsletter-subscribers.webp" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto/https%3A%2F%2Fcdn.broadwayinc.com%2Farticles%2Fimg%2Fnewsletter-subscribers.webp" alt="The Subscribers dialog listing addresses with a search box and paging" width="800" height="538"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;&lt;em&gt;The subscriber list, searchable by the start of an address.&lt;/em&gt;&lt;/p&gt;

&lt;p&gt;The &lt;strong&gt;Service Email&lt;/strong&gt; tab has no &lt;strong&gt;[Add]&lt;/strong&gt;. Its subscribers are your users, who subscribe and unsubscribe with their own account, so nobody can be added to it by hand.&lt;/p&gt;

&lt;h2&gt;
  
  
  Groups: several lists on one project
&lt;/h2&gt;

&lt;p&gt;One project is not limited to one newsletter. Click &lt;strong&gt;+ New group&lt;/strong&gt; to create a named list.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto/https%3A%2F%2Fcdn.broadwayinc.com%2Farticles%2Fimg%2Fnewsletter-group.webp" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto/https%3A%2F%2Fcdn.broadwayinc.com%2Farticles%2Fimg%2Fnewsletter-group.webp" alt="The New Group dialog with a name, a label and a Subscribers choice of Anyone, Any signed in user or Access group and above" width="800" height="538"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;&lt;em&gt;A named group: a short name that becomes part of its sending address, an optional label for the tab, and who may subscribe.&lt;/em&gt;&lt;/p&gt;

&lt;p&gt;The &lt;strong&gt;name&lt;/strong&gt; is two to twenty lowercase letters and digits. It becomes part of the group's sending address and cannot change later. The &lt;strong&gt;label&lt;/strong&gt; is what the tab shows. &lt;strong&gt;Subscribers&lt;/strong&gt; decides who may join: anyone with an email address, any signed-in user of your app, or only users at or above an access group you choose. That last option is how you run a members-only or paying-customers-only list without writing any code.&lt;/p&gt;

&lt;p&gt;Each group keeps its own subscribers, its own sending address and its own sent mail, so a subscriber of one never receives the other. A project can hold up to twenty of them.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto/https%3A%2F%2Fcdn.broadwayinc.com%2Farticles%2Fimg%2Fnewsletter-named-group.webp" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto/https%3A%2F%2Fcdn.broadwayinc.com%2Farticles%2Fimg%2Fnewsletter-named-group.webp" alt="The Promotions group tab showing its own sending address, its subscriber count, an Access row reading Anyone can subscribe, and a Delete group link" width="800" height="538"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;&lt;em&gt;A named group has its own address, its own count and an Access row. Deleting the group deletes its subscribers with it.&lt;/em&gt;&lt;/p&gt;

&lt;h2&gt;
  
  
  The signup form, without writing it
&lt;/h2&gt;

&lt;p&gt;The form is the part most people expect to build. You do not have to. Open your project's &lt;strong&gt;Settings&lt;/strong&gt; page and scroll to &lt;strong&gt;Application Examples&lt;/strong&gt;.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto/https%3A%2F%2Fcdn.broadwayinc.com%2Farticles%2Fimg%2Fnewsletter-examples.webp" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto/https%3A%2F%2Fcdn.broadwayinc.com%2Farticles%2Fimg%2Fnewsletter-examples.webp" alt="The Application Examples card on the Settings page, with a Newsletter row and an Open link" width="800" height="549"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;&lt;em&gt;Application Examples on the Settings page. Every row opens a plain HTML page that already runs against your project.&lt;/em&gt;&lt;/p&gt;

&lt;p&gt;Click &lt;strong&gt;[Open]&lt;/strong&gt; on the &lt;strong&gt;Newsletter&lt;/strong&gt; row. What opens is a working subscription page, already pointed at your project.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto/https%3A%2F%2Fcdn.broadwayinc.com%2Farticles%2Fimg%2Fnewsletter-form.webp" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto/https%3A%2F%2Fcdn.broadwayinc.com%2Farticles%2Fimg%2Fnewsletter-form.webp" alt="The example newsletter page: a short explanation, an email field and a Subscribe button" width="800" height="512"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;&lt;em&gt;The example page. Its address carries your project id, so it can be linked as it is.&lt;/em&gt;&lt;/p&gt;

&lt;p&gt;You can use it three ways:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Link it.&lt;/strong&gt; The address of that page carries your project id, so a link from your site, your Instagram bio or your email signature is a working signup form. Nothing to host.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Copy the form.&lt;/strong&gt; Click &lt;strong&gt;[Download]&lt;/strong&gt; on the same card for every example page as a ZIP, open &lt;code&gt;newsletter.html&lt;/code&gt;, and paste the form into your own page.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Write it.&lt;/strong&gt; It is a form with two fields, passed straight to the SDK:
&lt;/li&gt;
&lt;/ul&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight html"&gt;&lt;code&gt;&lt;span class="nt"&gt;&amp;lt;form&lt;/span&gt; &lt;span class="na"&gt;onsubmit=&lt;/span&gt;&lt;span class="s"&gt;"skapi.subscribeNewsletter(event).then(res =&amp;gt; alert(res))"&lt;/span&gt;&lt;span class="nt"&gt;&amp;gt;&lt;/span&gt;
    &lt;span class="nt"&gt;&amp;lt;input&lt;/span&gt; &lt;span class="na"&gt;type=&lt;/span&gt;&lt;span class="s"&gt;"email"&lt;/span&gt; &lt;span class="na"&gt;name=&lt;/span&gt;&lt;span class="s"&gt;"email"&lt;/span&gt; &lt;span class="na"&gt;placeholder=&lt;/span&gt;&lt;span class="s"&gt;"your@email.com"&lt;/span&gt; &lt;span class="nt"&gt;/&amp;gt;&lt;/span&gt;
    &lt;span class="nt"&gt;&amp;lt;input&lt;/span&gt; &lt;span class="na"&gt;hidden&lt;/span&gt; &lt;span class="na"&gt;name=&lt;/span&gt;&lt;span class="s"&gt;"group"&lt;/span&gt; &lt;span class="na"&gt;value=&lt;/span&gt;&lt;span class="s"&gt;"public"&lt;/span&gt; &lt;span class="nt"&gt;/&amp;gt;&lt;/span&gt;
    &lt;span class="nt"&gt;&amp;lt;input&lt;/span&gt; &lt;span class="na"&gt;hidden&lt;/span&gt; &lt;span class="na"&gt;name=&lt;/span&gt;&lt;span class="s"&gt;"redirect"&lt;/span&gt; &lt;span class="na"&gt;value=&lt;/span&gt;&lt;span class="s"&gt;"https://your.domain.com/thanks"&lt;/span&gt; &lt;span class="nt"&gt;/&amp;gt;&lt;/span&gt;
    &lt;span class="nt"&gt;&amp;lt;input&lt;/span&gt; &lt;span class="na"&gt;type=&lt;/span&gt;&lt;span class="s"&gt;"submit"&lt;/span&gt; &lt;span class="na"&gt;value=&lt;/span&gt;&lt;span class="s"&gt;"Subscribe"&lt;/span&gt; &lt;span class="nt"&gt;/&amp;gt;&lt;/span&gt;
&lt;span class="nt"&gt;&amp;lt;/form&amp;gt;&lt;/span&gt;
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;Change &lt;code&gt;public&lt;/code&gt; to a group's name to feed a named group instead, and &lt;code&gt;redirect&lt;/code&gt; is where the visitor lands after confirming.&lt;/p&gt;

&lt;p&gt;A visitor who subscribes this way gets a confirmation email and joins the list when they click the link in it, so every address on your public list is one that proved it wanted to be there. A signed-in user with a verified email skips the confirmation.&lt;/p&gt;

&lt;h2&gt;
  
  
  Rules that keep you out of the spam folder
&lt;/h2&gt;

&lt;p&gt;The plumbing that a mailing tool charges for is switched on for every project:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Confirmation&lt;/strong&gt; for every public signup, and an &lt;strong&gt;unsubscribe link&lt;/strong&gt; on every newsletter.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Complaint shutoff.&lt;/strong&gt; A project that collects five spam complaints in a month, or a complaint rate of half a percent once a hundred mails have gone out, stops sending newsletters. Nothing else about the project stops, and we clear the block after a look.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Sender trust.&lt;/strong&gt; Only mail from your registered address, passing SPF, DKIM and DMARC, is forwarded.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Counted per recipient.&lt;/strong&gt; One newsletter to a list of 1,000 uses 1,000 of your monthly sends, so the number on your plan is a number of deliveries, not of campaigns.&lt;/li&gt;
&lt;/ul&gt;

&lt;h2&gt;
  
  
  What it costs
&lt;/h2&gt;

&lt;p&gt;Newsletters are part of every plan, not a separate product.&lt;/p&gt;

&lt;div class="table-wrapper-paragraph"&gt;&lt;table&gt;
&lt;thead&gt;
&lt;tr&gt;
&lt;th&gt;&lt;/th&gt;
&lt;th&gt;Free&lt;/th&gt;
&lt;th&gt;Standard $19/mo&lt;/th&gt;
&lt;th&gt;Premium $89/mo&lt;/th&gt;
&lt;/tr&gt;
&lt;/thead&gt;
&lt;tbody&gt;
&lt;tr&gt;
&lt;td&gt;Subscribers, all groups&lt;/td&gt;
&lt;td&gt;50&lt;/td&gt;
&lt;td&gt;5,000&lt;/td&gt;
&lt;td&gt;50,000&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Newsletter sends per month&lt;/td&gt;
&lt;td&gt;50&lt;/td&gt;
&lt;td&gt;5,000&lt;/td&gt;
&lt;td&gt;50,000&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Email storage&lt;/td&gt;
&lt;td&gt;50 MB&lt;/td&gt;
&lt;td&gt;1 GB&lt;/td&gt;
&lt;td&gt;5 GB&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Past a limit&lt;/td&gt;
&lt;td&gt;Stop&lt;/td&gt;
&lt;td&gt;Stop&lt;/td&gt;
&lt;td&gt;$1 per 1,000 extra sends or subscribers&lt;/td&gt;
&lt;/tr&gt;
&lt;/tbody&gt;
&lt;/table&gt;&lt;/div&gt;

&lt;p&gt;The same plan carries your database, file storage, hosting, user accounts, automated emails and everything else a Skapi project does. The newsletter is not the thing you are paying for. It comes with it.&lt;/p&gt;

&lt;p&gt;Now Mailchimp, from its pricing page at the time of writing. Its Free plan takes 250 contacts and 500 sends a month. The paid plans start at $13 a month for Essentials and $20 for Standard, both for up to 500 contacts, and the price climbs with every tier of contacts you hold. Premium starts at $350.&lt;/p&gt;

&lt;div class="table-wrapper-paragraph"&gt;&lt;table&gt;
&lt;thead&gt;
&lt;tr&gt;
&lt;th&gt;Contacts&lt;/th&gt;
&lt;th&gt;Mailchimp Essentials&lt;/th&gt;
&lt;th&gt;Mailchimp Standard&lt;/th&gt;
&lt;/tr&gt;
&lt;/thead&gt;
&lt;tbody&gt;
&lt;tr&gt;
&lt;td&gt;500&lt;/td&gt;
&lt;td&gt;$13&lt;/td&gt;
&lt;td&gt;$20&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;2,500&lt;/td&gt;
&lt;td&gt;$45&lt;/td&gt;
&lt;td&gt;$60&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;5,000&lt;/td&gt;
&lt;td&gt;$75&lt;/td&gt;
&lt;td&gt;$100&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;10,000&lt;/td&gt;
&lt;td&gt;$110&lt;/td&gt;
&lt;td&gt;$135&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;25,000&lt;/td&gt;
&lt;td&gt;$270&lt;/td&gt;
&lt;td&gt;$310&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;50,000&lt;/td&gt;
&lt;td&gt;$385&lt;/td&gt;
&lt;td&gt;$450&lt;/td&gt;
&lt;/tr&gt;
&lt;/tbody&gt;
&lt;/table&gt;&lt;/div&gt;

&lt;p&gt;Put the two side by side for a monthly newsletter, one send to the whole list each month:&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2F5kcbktc3tyzojkh4idrp.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2F5kcbktc3tyzojkh4idrp.png" alt="Mailchimp cheaper alternative" width="800" height="483"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;$19 for 5,000 subscribers, against $75 to $100. $89 for 50,000, against $385 to $450.&lt;/strong&gt;&lt;/p&gt;

&lt;p&gt;&lt;em&gt;The same list at both companies, at the time of writing. Skapi's price is the plan the list fits in, Standard up to 5,000 subscribers and Premium above it, and that plan runs the rest of your backend too. Mailchimp's is its published tier for that many contacts.&lt;/em&gt;&lt;/p&gt;

&lt;p&gt;Two honest notes. Below a few hundred subscribers, Mailchimp's entry tier is a few dollars cheaper than our Standard plan, and its free tier is larger than ours. And Skapi counts sends per recipient, so a weekly newsletter to 5,000 people is 20,000 sends a month, which puts you on Premium at $89, about what Mailchimp charges for the same list. At 25,000 subscribers sent weekly, Skapi is $139 with the extra sends billed, against $270 and up. The bigger the list, the wider the gap.&lt;/p&gt;

&lt;p&gt;One more difference in how the count works. Mailchimp's tiers are based on the contacts in your audience, which by its own rules includes people who have unsubscribed until you archive them. Skapi counts subscribers. When someone unsubscribes, they leave the list and the count.&lt;/p&gt;

&lt;h2&gt;
  
  
  What Mailchimp has that we do not
&lt;/h2&gt;

&lt;p&gt;Mailchimp is a marketing suite. It has a drag-and-drop editor, a template gallery, automation journeys, A/B tests, landing pages and deep reporting. If your business runs on those, keep it.&lt;/p&gt;

&lt;p&gt;Skapi gives you a list, a form, an address, and a table that tells you who read, who complained and who bounced. You write in your own mail app, and segmentation is a group. For a great many businesses, that is the entire newsletter they ever wanted, and it is the part that Mailchimp charges for by the contact.&lt;/p&gt;

&lt;h2&gt;
  
  
  Where to start
&lt;/h2&gt;

&lt;ol&gt;
&lt;li&gt;Register an email alias on your project's Settings page.&lt;/li&gt;
&lt;li&gt;Open &lt;strong&gt;Newsletters&lt;/strong&gt;, add the subscribers you already have with &lt;strong&gt;[Add]&lt;/strong&gt;, and create a group if you need more than one list.&lt;/li&gt;
&lt;li&gt;Link the example signup page, or paste its form into your site.&lt;/li&gt;
&lt;li&gt;Write the newsletter in your mail app and send it to the address on the page.&lt;/li&gt;
&lt;/ol&gt;

&lt;p&gt;The first one takes about ten minutes. Every one after that takes as long as writing the email.&lt;/p&gt;

</description>
      <category>productivity</category>
      <category>automation</category>
      <category>startup</category>
      <category>tools</category>
    </item>
    <item>
      <title>How to cut the token bill of vibe coding: Skapi, backend built for AI agents</title>
      <dc:creator>Skapi</dc:creator>
      <pubDate>Mon, 12 Oct 2026 06:34:46 +0000</pubDate>
      <link>https://dev.to/skapi_api/how-to-cut-the-token-bill-of-vibe-coding-skapi-backend-built-for-ai-agents-33jm</link>
      <guid>https://dev.to/skapi_api/how-to-cut-the-token-bill-of-vibe-coding-skapi-backend-built-for-ai-agents-33jm</guid>
      <description>&lt;h2&gt;
  
  
  Cut the token bill of vibe coding: Skapi, backend built for AI agents
&lt;/h2&gt;

&lt;p&gt;Every prompt you send an AI coding agent costs tokens, and a surprising share of them go to the backend: schemas, policies, generated types, redeploys. We measured where the tokens go when the backend is built through an MCP server on Supabase or Firebase, why the spending never stops after the first build, and what the same app costs when it runs on skapi-js instead. We measured our own side the same way: the prompt file the agent loads, the pages it reads, and the Skapi MCP server. The short version: Skapi is the backend that costs your agent the least, because there is no schema to rebuild, ever.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fcdn.broadwayinc.com%2Fimg%2Fskapi-logo-cut-small.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fcdn.broadwayinc.com%2Fimg%2Fskapi-logo-cut-small.png" alt="Skapi" width="244" height="300"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;h2&gt;
  
  
  Where the tokens go
&lt;/h2&gt;

&lt;p&gt;An AI coding agent works in round trips. It reads its context, calls a tool, reads the result, and calls again. Every one of those trips sends the whole context back through the model: the system prompt, the instruction file your tool loads from the project root, the conversation so far, and the description of every tool the agent is allowed to use. Prompt caching makes the repeated part cheaper, usually to about a tenth of the price, but it never makes it free, and cache entries expire in minutes.&lt;/p&gt;

&lt;p&gt;Build a backend through an MCP server and the loop looks like this. The agent lists the tables to learn the schema, writes a migration, applies it, writes a row-level security policy per operation per table, regenerates the TypeScript types, runs the security advisor, fixes what it reports, and only then edits the client code. Then you ask for one more field, and the loop runs again from the migration step.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Flve0gmnfjigwe20jnjfw.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Flve0gmnfjigwe20jnjfw.png" alt="backend through MCP" width="800" height="261"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;&lt;em&gt;The loop an agent runs for every schema change, and the loop it runs on skapi-js. The steps in the first row are the tools Supabase's MCP server describes for exactly this purpose.&lt;/em&gt;&lt;/p&gt;

&lt;p&gt;Two things make the first loop expensive, and neither shows up on a pricing page. The steps are round trips, so the growing context is re-read at each one. And the artifacts are not small: a migration is hundreds of tokens, the policies are hundreds more, and the regenerated types file is re-read in full after every change.&lt;/p&gt;

&lt;h2&gt;
  
  
  We measured it
&lt;/h2&gt;

&lt;p&gt;The numbers come from the real servers and the real files. We started the &lt;code&gt;@supabase/mcp-server-supabase&lt;/code&gt; (0.13.0) and &lt;code&gt;firebase-tools&lt;/code&gt; (15.31.0) packages over stdio, sent &lt;code&gt;tools/list&lt;/code&gt;, and counted the payload the agent receives with the cl100k tokenizer. We counted Supabase's published AI prompt files the same way, from the &lt;code&gt;examples/prompts&lt;/code&gt; folder of its repository on 2026-10-07, and on the Skapi side the prompt file, the documentation pages as published that day, and our own MCP server's &lt;code&gt;tools/list&lt;/code&gt;. Claude's tokenizer differs a little, so read these as sizes, not invoices. The migration, policies and types on the Supabase side are our own minimal examples for a task app with &lt;code&gt;profiles&lt;/code&gt;, &lt;code&gt;projects&lt;/code&gt; and &lt;code&gt;tasks&lt;/code&gt;, written the way Supabase's own guides say to write them.&lt;/p&gt;

&lt;h3&gt;
  
  
  What travels with every request
&lt;/h3&gt;

&lt;p&gt;Two things ride along on every round trip for as long as the session lasts: the tool list an MCP server describes to the model, and the instruction file your coding tool loads from the project root (&lt;code&gt;CLAUDE.md&lt;/code&gt;, &lt;code&gt;AGENTS.md&lt;/code&gt;, &lt;code&gt;.cursorrules&lt;/code&gt; and the rest). Supabase's tool list alone is 5,051 tokens in its default configuration, and 1,051 with only the database and development tools, but its documentation tells you to add its prompt files to that instruction file, for RLS policies, migrations, the declarative schema, the SQL style guide and the auth bootstrap, because the agent writes them wrong without the rules. Those eight files are 12,563 tokens; the five a task app needs are 6,552. On the Skapi side there is no server in the build loop, so the standing cost is the instruction file alone.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fe2dhz5apkz9rafh09qm6.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fe2dhz5apkz9rafh09qm6.png" alt="Skapi is serverless" width="799" height="284"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;&lt;em&gt;Compact JSON as the client receives it, plus the prompt files each side tells you to load. Firebase's number is large because its Firestore document tools carry their full JSON schemas. The Supabase rows are its default tool list plus its own published prompt files, all eight or the five a task app needs. The two Skapi rows are the same session: 5,847 tokens of prompt file, and no tool list, because the agent builds against the SDK, not through a server.&lt;/em&gt;&lt;/p&gt;

&lt;p&gt;Firebase's &lt;code&gt;--only&lt;/code&gt; flag and Supabase's &lt;code&gt;features&lt;/code&gt; parameter cut those lists down, and you should use them. But the smallest Supabase set that can still apply a migration is about a thousand tokens, most people run the default, and the prompt files do not shrink with the flag. Take a modest working session of 40 tool calls. On Supabase, with the default list and the prompt files a task app needs, that is about 460,000 tokens of standing context passing through the model. On Firebase's default list it is 2.4 million before any instruction file. On Skapi it is about 230,000, and that is everything: the rules, the map and the whole method index.&lt;/p&gt;

&lt;p&gt;The Skapi row is the whole instruction file: &lt;code&gt;SKAPI.md&lt;/code&gt; is 24.7 KB, 5,847 tokens, and it holds the rules, a map of the documentation and an index of every SDK method, each linking to its page. It is sized to be loaded on every turn by every tool, Codex CLI's 32 KiB limit on &lt;code&gt;AGENTS.md&lt;/code&gt; included, and the pages themselves are read only when a feature needs them, which is the next number.&lt;/p&gt;

&lt;h3&gt;
  
  
  The change loop, three steps
&lt;/h3&gt;

&lt;p&gt;Now the artifacts. We built the same task app three times over: the first build, then "add a priority to tasks and filter by it", then "let me share a task with a teammate". For each step we counted what the agent has to write for the backend and what it has to read back.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fbtitwdfhroq1t1lxbxay.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fbtitwdfhroq1t1lxbxay.png" alt="Artefacts" width="800" height="469"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;&lt;em&gt;Each Supabase bar is what one step costs in backend artifacts: the first segment is what the agent writes, the second is the regenerated types file it reads back afterwards. Three steps come to 6,705 tokens, and 5,253 of them, 78 percent, are the same types file read back three times. The skapi-js rows have no bar to draw.&lt;/em&gt;&lt;/p&gt;

&lt;div class="table-wrapper-paragraph"&gt;&lt;table&gt;
&lt;thead&gt;
&lt;tr&gt;
&lt;th&gt;Step&lt;/th&gt;
&lt;th&gt;Supabase over MCP&lt;/th&gt;
&lt;th&gt;skapi-js&lt;/th&gt;
&lt;/tr&gt;
&lt;/thead&gt;
&lt;tbody&gt;
&lt;tr&gt;
&lt;td&gt;
&lt;strong&gt;Build&lt;/strong&gt;: accounts, tasks with an attachment&lt;/td&gt;
&lt;td&gt;Migration with 3 tables, 3 indexes, RLS on every table and 12 policies: 787 tokens. Storage bucket and 3 object policies: 159. Regenerated types, read back: 1,751. &lt;strong&gt;2,697 tokens&lt;/strong&gt;, plus &lt;code&gt;list_tables&lt;/code&gt; and &lt;code&gt;get_advisors&lt;/code&gt; responses we could not measure without a live project.&lt;/td&gt;
&lt;td&gt;
&lt;strong&gt;0&lt;/strong&gt;. Accounts come with the project. A record carries its own table name, access group and index, so there is nothing to create before the first &lt;code&gt;postRecord()&lt;/code&gt;.&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;
&lt;strong&gt;Change 1&lt;/strong&gt;: add a priority, filter by it&lt;/td&gt;
&lt;td&gt;Migration adding the column and an index: 38. Regenerated types, read back in full: 1,751. &lt;strong&gt;1,789 tokens&lt;/strong&gt;
&lt;/td&gt;
&lt;td&gt;
&lt;strong&gt;0&lt;/strong&gt;. The change is an &lt;code&gt;index&lt;/code&gt; key in the call that already exists: about 100 tokens of client edit.&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;
&lt;strong&gt;Change 2&lt;/strong&gt;: share a task with a teammate&lt;/td&gt;
&lt;td&gt;New &lt;code&gt;task_shares&lt;/code&gt; table, its RLS, and the two task policies rewritten with &lt;code&gt;exists&lt;/code&gt; subqueries: 468. Regenerated types: 1,751. &lt;strong&gt;2,219 tokens&lt;/strong&gt;
&lt;/td&gt;
&lt;td&gt;
&lt;strong&gt;0&lt;/strong&gt;. One call: &lt;code&gt;skapi.grantPrivateRecordAccess({ record_id, user_id })&lt;/code&gt;, 28 tokens.&lt;/td&gt;
&lt;/tr&gt;
&lt;/tbody&gt;
&lt;/table&gt;&lt;/div&gt;

&lt;p&gt;The client code is roughly the same shape on both sides. Our supabase-js module for the build step is 408 tokens, the skapi-js one 202. The difference is not the client. It is everything the client needs to exist first.&lt;/p&gt;

&lt;p&gt;Notice the pattern in the middle column. The migration for change 1 is 38 tokens. The regenerated types the agent then reads back are 1,751, and they are 1,751 again after change 2, and after every change after that, because the tool returns the whole file. The steady cost of a schema is not the schema. It is the re-reading.&lt;/p&gt;

&lt;p&gt;And these are the artifacts alone. Each of those steps is a tool call, and each tool call is a round trip that re-reads the conversation and the tool list. A change that costs 1,789 tokens of artifacts costs several times that by the time the agent has listed the tables, applied the migration, regenerated the types, run the advisor and read every response.&lt;/p&gt;

&lt;h2&gt;
  
  
  What the agent reads instead
&lt;/h2&gt;

&lt;p&gt;The fair question is what the agent reads on the Skapi side, since it is not reading a tool list. It reads documentation, once per page, when the feature comes up. The prompt file tells it which page to read: every page of the documentation is published as raw markdown under &lt;code&gt;docs.skapi.com/md/&lt;/code&gt;, 94 pages, and the file links to the 71 on the sidebar; the rest are reached through the links between pages.&lt;/p&gt;

&lt;p&gt;For the task app above the agent reads ten guide pages: getting started, signing up and logging in, creating, fetching, updating and deleting records, access restrictions, files and indexing. Measured on the published files, they come to 25,401 tokens. The reference entries it is told to check before writing a call live on three more pages, authentication, database and data types, 23,718 tokens if it reads all three in full, and the data types page is 11,280 of that. So the most an agent reads for this app is about 49,000 tokens, read once, and a careful agent that reads the entries it needs rather than whole pages reads less. The median sidebar page is 2,154 tokens.&lt;/p&gt;

&lt;p&gt;The difference is what happens after the first read. On Skapi, nothing in the documentation changes when your app changes, so it is never re-read. Every schema change on the other side produces a new artifact that the agent has to read back. An agent building on Supabase reads its guides too; the RLS guide alone is the one that tells it to write a policy per operation per table. We did not measure those reads, so they are not in the charts, but they are not zero either.&lt;/p&gt;

&lt;p&gt;Put the session together and it reads like this. On Skapi: 5,847 tokens of prompt file on every turn, the pages the app needs once, and no backend artifact, ever. Through an MCP server: the tool list on every turn, the guides once, and a migration, a policy set and a regenerated types file on every change.&lt;/p&gt;

&lt;p&gt;Skapi's prompt file is what the dashboard downloads under the name your tool expects, &lt;code&gt;CLAUDE.md&lt;/code&gt;, &lt;code&gt;AGENTS.md&lt;/code&gt;, &lt;code&gt;.cursorrules&lt;/code&gt; and the rest. Open your project's Settings page and look for &lt;strong&gt;For AI Agents&lt;/strong&gt;.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto/https%3A%2F%2Fcdn.broadwayinc.com%2Farticles%2Fimg%2Fvibe-agents-card.webp" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto/https%3A%2F%2Fcdn.broadwayinc.com%2Farticles%2Fimg%2Fvibe-agents-card.webp" alt="The For AI Agents card on a project's Settings page: the opening prompt with the project ID filled in, and a Download row per tool from Claude Code to Windsurf" width="800" height="596"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;&lt;em&gt;The For AI Agents card. Click the prompt to copy it, then download the file under the name your agent reads.&lt;/em&gt;&lt;/p&gt;

&lt;h2&gt;
  
  
  And the Skapi MCP server
&lt;/h2&gt;

&lt;p&gt;Skapi has an MCP server too, and it is useful: it lets an agent manage your project, read and write records, register tickets, store secret keys and set up OpenID login, signed in as you. Measured the same way as the others, its public &lt;code&gt;tools/list&lt;/code&gt; is 17 tools and 7,610 tokens, in the same range as Supabase's default list. It is simply not part of the build loop. The agent writes code against &lt;code&gt;skapi-js&lt;/code&gt;, and the backend answers; the server is for the moments when you want the agent to operate the project rather than write the app. It carries no documentation of its own: when an agent on a platform that cannot fetch urls needs a page, the server's &lt;code&gt;readDocs&lt;/code&gt; tool returns that one page from the same source.&lt;/p&gt;

&lt;h2&gt;
  
  
  When the meter is credits instead of tokens
&lt;/h2&gt;

&lt;p&gt;Web builders such as Lovable do not show you tokens. They charge credits per message, and the price of a message depends on how much work the agent did. At the time of writing, Lovable's own examples put "Make the button gray" at 0.50 credits and "Add authentication with sign up and login" at 1.20, the free plan grants 5 build credits a day, and the entry Pro tier is $25 for 100 credits a month, so a message that makes the agent do more costs you more of them.&lt;/p&gt;

&lt;p&gt;The meter is different, the mechanism is not. A message that has to rebuild a schema, rewrite policies and redeploy is a message that does more work, and it costs more credits than one that edits a line of client code. The backend you choose decides what "add a field" means to the agent, and therefore what it costs you, on every platform that charges for the agent's effort.&lt;/p&gt;

&lt;h2&gt;
  
  
  Why Skapi is built for AI agents
&lt;/h2&gt;

&lt;p&gt;An agent is at its best when the whole backend is one SDK it can write against, with nothing to provision, migrate or deploy in between. That is what Skapi is: accounts, database, files, CDN, emails, realtime, hosting and webhooks behind one JavaScript library, on a plain HTML page or in any framework, with a free plan to start on.&lt;/p&gt;

&lt;p&gt;Skapi has no schema. A record is a piece of data with a few words attached: the table it belongs to, who may read it, and optionally what it is indexed by. The table exists because a record said so. That is what makes the agent's side so small: there is no relational schema to design before the first record, no policy language to learn, and nothing to deploy, so the design work an agent does is the one word that says who may read the record.&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight javascript"&gt;&lt;code&gt;&lt;span class="nx"&gt;skapi&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;postRecord&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="nx"&gt;form&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="p"&gt;{&lt;/span&gt;
    &lt;span class="na"&gt;table&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="p"&gt;{&lt;/span&gt; &lt;span class="na"&gt;name&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="s1"&gt;tasks&lt;/span&gt;&lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="na"&gt;access_group&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="s1"&gt;private&lt;/span&gt;&lt;span class="dl"&gt;'&lt;/span&gt; &lt;span class="p"&gt;},&lt;/span&gt;
    &lt;span class="na"&gt;index&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="p"&gt;{&lt;/span&gt; &lt;span class="na"&gt;name&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="s1"&gt;priority&lt;/span&gt;&lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="na"&gt;value&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="mi"&gt;2&lt;/span&gt; &lt;span class="p"&gt;},&lt;/span&gt;
    &lt;span class="na"&gt;reference&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="nx"&gt;projectId&lt;/span&gt;
&lt;span class="p"&gt;});&lt;/span&gt;
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;That is the whole backend side of a task with an attachment. &lt;code&gt;form&lt;/code&gt; is the HTML form the user filled in, file input included, and the file is stored and served from a CDN without another line. &lt;code&gt;access_group: 'private'&lt;/code&gt; is the entire security policy: the record's owner can read it, nobody else can, until the owner grants someone access. Change &lt;code&gt;'private'&lt;/code&gt; to &lt;code&gt;'authorized'&lt;/code&gt; and every signed-in user can read it; change it to &lt;code&gt;'public'&lt;/code&gt; and everyone can.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto/https%3A%2F%2Fcdn.broadwayinc.com%2Farticles%2Fimg%2Fvibe-records.webp" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto/https%3A%2F%2Fcdn.broadwayinc.com%2Farticles%2Fimg%2Fvibe-records.webp" alt="The Database page of a Skapi project: records from the tasks, projects, comments, attachments, profiles and settings tables in one list, each with an access icon, an ID, tags and its data" width="800" height="538"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;&lt;em&gt;The Database page after the three steps above. Six tables, none of them created by hand. The lock, group and globe icons are the access groups: private, authorized and public.&lt;/em&gt;&lt;/p&gt;

&lt;p&gt;Open any record and you are looking at the entire data model of a Skapi project: the table name, the access group, one index, a reference, tags, and the data itself.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto/https%3A%2F%2Fcdn.broadwayinc.com%2Farticles%2Fimg%2Fvibe-record.webp" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto/https%3A%2F%2Fcdn.broadwayinc.com%2Farticles%2Fimg%2Fvibe-record.webp" alt="A single task record opened on the Database page: record info, table name and access group, subscription flags, the priority index, options, the JSON data and a Files section" width="800" height="975"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;&lt;em&gt;One record. Everything an agent would otherwise put into a migration is a field on this form, and every field is optional except the table name.&lt;/em&gt;&lt;/p&gt;

&lt;p&gt;Accounts work the same way. &lt;code&gt;skapi.signup(event)&lt;/code&gt; on a form creates the user, &lt;code&gt;skapi.login(event)&lt;/code&gt; signs them in, and the confirmation email, the password reset, the welcome mail and the user list on your dashboard are already there.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto/https%3A%2F%2Fcdn.broadwayinc.com%2Farticles%2Fimg%2Fvibe-users.webp" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto/https%3A%2F%2Fcdn.broadwayinc.com%2Farticles%2Fimg%2Fvibe-users.webp" alt="The Users page of a Skapi project: five users with email, name and approval status, and toolbar actions to grant access, block, unblock and delete" width="800" height="538"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;&lt;em&gt;The Users page. Signup, login, verification, password reset and the admin actions come with the project.&lt;/em&gt;&lt;/p&gt;

&lt;p&gt;The same project also carries what a growing app asks for next, none of it a separate build: realtime messaging, automated emails and newsletters, secret keys and request forwarding for third-party APIs, tickets that accept webhooks, OpenID login and web hosting. Each one is a method on the SDK or a card on the dashboard, and each one is one loop of the second kind.&lt;/p&gt;

&lt;h2&gt;
  
  
  Where to start
&lt;/h2&gt;

&lt;p&gt;The cheapest way to check any of this is to build something. A Free project costs nothing.&lt;/p&gt;

&lt;ol&gt;
&lt;li&gt;Create a project at &lt;a href="https://www.skapi.com" rel="noopener noreferrer"&gt;skapi.com&lt;/a&gt; and copy its Project ID from the Settings page.&lt;/li&gt;
&lt;li&gt;On the same page, download the prompt file for your tool from &lt;strong&gt;For AI Agents&lt;/strong&gt;, and put it in the root of your project.&lt;/li&gt;
&lt;li&gt;Start the agent with the opening prompt: &lt;code&gt;My Skapi project ID is "&amp;lt;Project ID&amp;gt;". Build me a task board with accounts, private tasks and attachments.&lt;/code&gt;
&lt;/li&gt;
&lt;li&gt;Ask for the next change. Then the one after that.&lt;/li&gt;
&lt;/ol&gt;

&lt;p&gt;The first build is an afternoon. Every change after it costs what a change to the client code costs, because that is all it is. Your agent's tokens go into your app, not into the plumbing under it.&lt;/p&gt;

</description>
      <category>ai</category>
      <category>vibecoding</category>
      <category>programming</category>
      <category>productivity</category>
    </item>
    <item>
      <title>The real price of a backend in São Paulo: $529 on AWS vs $19 on Skapi</title>
      <dc:creator>Skapi</dc:creator>
      <pubDate>Fri, 02 Oct 2026 05:52:52 +0000</pubDate>
      <link>https://dev.to/skapi_api/the-real-price-of-a-backend-in-sao-paulo-529-on-aws-vs-19-on-skapi-l3e</link>
      <guid>https://dev.to/skapi_api/the-real-price-of-a-backend-in-sao-paulo-529-on-aws-vs-19-on-skapi-l3e</guid>
      <description>&lt;h2&gt;
  
  
  The real price of a backend in São Paulo: $529 on AWS vs $19 on Skapi
&lt;/h2&gt;

&lt;p&gt;In &lt;a href="https://www.skapi.com/articles/2026-09-28-00" rel="noopener noreferrer"&gt;the last article&lt;/a&gt; we said a Skapi project in São Paulo runs at a fraction of what a traditional server backend costs there, and that getting there was hard. Here is the bill, line by line: $528.71 a month for the traditional stack, $19 for the same backend on Skapi, and how we made that work.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fcdn.broadwayinc.com%2Fimg%2Fskapi-logo-cut-small.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fcdn.broadwayinc.com%2Fimg%2Fskapi-logo-cut-small.png" alt="Skapi" width="244" height="300"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;h2&gt;
  
  
  Why São Paulo is expensive
&lt;/h2&gt;

&lt;p&gt;AWS charges more in São Paulo than in most of its regions, and it is not a rounding error. Here is what the same building blocks cost in São Paulo (sa-east-1) and in N. Virginia (us-east-1), the region most tutorials silently assume you are in. List prices, on demand, as of September 2026.&lt;/p&gt;

&lt;div class="table-wrapper-paragraph"&gt;&lt;table&gt;
&lt;thead&gt;
&lt;tr&gt;
&lt;th&gt;Building block&lt;/th&gt;
&lt;th&gt;São Paulo&lt;/th&gt;
&lt;th&gt;N. Virginia&lt;/th&gt;
&lt;th&gt;São Paulo costs&lt;/th&gt;
&lt;/tr&gt;
&lt;/thead&gt;
&lt;tbody&gt;
&lt;tr&gt;
&lt;td&gt;App server, t3.medium, per hour&lt;/td&gt;
&lt;td&gt;$0.0672&lt;/td&gt;
&lt;td&gt;$0.0416&lt;/td&gt;
&lt;td&gt;62% more&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Database, RDS db.t3.medium Multi-AZ, per hour&lt;/td&gt;
&lt;td&gt;$0.302&lt;/td&gt;
&lt;td&gt;$0.145&lt;/td&gt;
&lt;td&gt;108% more&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Database storage, gp3, per GB-month&lt;/td&gt;
&lt;td&gt;$0.219&lt;/td&gt;
&lt;td&gt;$0.115&lt;/td&gt;
&lt;td&gt;90% more&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Server disk, EBS gp3, per GB-month&lt;/td&gt;
&lt;td&gt;$0.152&lt;/td&gt;
&lt;td&gt;$0.080&lt;/td&gt;
&lt;td&gt;90% more&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;NAT gateway, per hour&lt;/td&gt;
&lt;td&gt;$0.093&lt;/td&gt;
&lt;td&gt;$0.045&lt;/td&gt;
&lt;td&gt;107% more&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Application load balancer, per hour&lt;/td&gt;
&lt;td&gt;$0.034&lt;/td&gt;
&lt;td&gt;$0.0225&lt;/td&gt;
&lt;td&gt;51% more&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;File storage, S3 Standard, per GB-month&lt;/td&gt;
&lt;td&gt;$0.0405&lt;/td&gt;
&lt;td&gt;$0.023&lt;/td&gt;
&lt;td&gt;76% more&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Data transfer out, per GB&lt;/td&gt;
&lt;td&gt;$0.150&lt;/td&gt;
&lt;td&gt;$0.090&lt;/td&gt;
&lt;td&gt;67% more&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;CDN delivery, CloudFront, per GB&lt;/td&gt;
&lt;td&gt;$0.110&lt;/td&gt;
&lt;td&gt;$0.085&lt;/td&gt;
&lt;td&gt;29% more&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;CDN requests, CloudFront HTTPS, per 10,000&lt;/td&gt;
&lt;td&gt;$0.022&lt;/td&gt;
&lt;td&gt;$0.010&lt;/td&gt;
&lt;td&gt;120% more&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Email, SES, per 1,000 messages&lt;/td&gt;
&lt;td&gt;$0.10&lt;/td&gt;
&lt;td&gt;$0.10&lt;/td&gt;
&lt;td&gt;the same&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;User accounts, Cognito Lite, per monthly active user&lt;/td&gt;
&lt;td&gt;$0.0055&lt;/td&gt;
&lt;td&gt;$0.0055&lt;/td&gt;
&lt;td&gt;the same&lt;/td&gt;
&lt;/tr&gt;
&lt;/tbody&gt;
&lt;/table&gt;&lt;/div&gt;

&lt;p&gt;CDN prices follow the viewer, not the origin, so a Brazilian audience pays the South America rate wherever the files live.&lt;/p&gt;

&lt;p&gt;Two things stand out. The parts that run all month, servers, databases and gateways, carry the steepest premium. The parts that only cost when someone uses them, email and identity, cost the same everywhere. Keep that split in mind. It is the whole story.&lt;/p&gt;

&lt;h2&gt;
  
  
  The bill for a traditional backend
&lt;/h2&gt;

&lt;p&gt;Let's price a modest production app. Not a toy: 50,000 registered users, 20,000 of whom sign in each month, 3 GB of database, 120 GB of uploaded files, 100 GB served through the CDN each month, and 5,000 transactional emails. That is the exact ceiling of Skapi's Standard plan, so the comparison at the end is fair.&lt;/p&gt;

&lt;p&gt;Built the traditional way in São Paulo, with the redundancy you would want for something real, the bill looks like this:&lt;/p&gt;

&lt;div class="table-wrapper-paragraph"&gt;&lt;table&gt;
&lt;thead&gt;
&lt;tr&gt;
&lt;th&gt;Line&lt;/th&gt;
&lt;th&gt;What it is&lt;/th&gt;
&lt;th&gt;Per month&lt;/th&gt;
&lt;/tr&gt;
&lt;/thead&gt;
&lt;tbody&gt;
&lt;tr&gt;
&lt;td&gt;App servers&lt;/td&gt;
&lt;td&gt;2 × t3.medium, one per availability zone&lt;/td&gt;
&lt;td&gt;$98.11&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Server disks&lt;/td&gt;
&lt;td&gt;2 × 30 GB gp3, plus daily snapshots&lt;/td&gt;
&lt;td&gt;$13.20&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Load balancer&lt;/td&gt;
&lt;td&gt;Application Load Balancer, one capacity unit&lt;/td&gt;
&lt;td&gt;$32.85&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;NAT gateway&lt;/td&gt;
&lt;td&gt;Outbound access for servers in private subnets&lt;/td&gt;
&lt;td&gt;$69.75&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Database&lt;/td&gt;
&lt;td&gt;RDS PostgreSQL db.t3.medium, Multi-AZ, 20 GB&lt;/td&gt;
&lt;td&gt;$229.22&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;File storage&lt;/td&gt;
&lt;td&gt;S3, 120 GB and its requests&lt;/td&gt;
&lt;td&gt;$6.68&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;CDN&lt;/td&gt;
&lt;td&gt;CloudFront, 100 GB and 2 million requests&lt;/td&gt;
&lt;td&gt;$15.40&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;API traffic out&lt;/td&gt;
&lt;td&gt;30 GB of responses to the internet&lt;/td&gt;
&lt;td&gt;$4.50&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Email&lt;/td&gt;
&lt;td&gt;SES, 5,000 messages&lt;/td&gt;
&lt;td&gt;$0.50&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;User accounts&lt;/td&gt;
&lt;td&gt;Cognito Lite, 20,000 active users, first 10,000 free&lt;/td&gt;
&lt;td&gt;$55.00&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;DNS and monitoring&lt;/td&gt;
&lt;td&gt;Route 53 zone, CloudWatch basics&lt;/td&gt;
&lt;td&gt;$3.50&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;&lt;strong&gt;Total&lt;/strong&gt;&lt;/td&gt;
&lt;td&gt;&lt;/td&gt;
&lt;td&gt;&lt;strong&gt;$528.71&lt;/strong&gt;&lt;/td&gt;
&lt;/tr&gt;
&lt;/tbody&gt;
&lt;/table&gt;&lt;/div&gt;

&lt;p&gt;You can cut that down. Run one server in a public subnet, drop the load balancer and the NAT gateway, run the database in a single availability zone, and it comes to about $256 a month. It also means the site goes down every time you deploy, reboot, or the database enters its maintenance window, and there is nothing to fail over to when the one server dies.&lt;/p&gt;

&lt;p&gt;For scale: the same production stack in N. Virginia comes to about $311. São Paulo adds 70% to the bill for the privilege of being close to your users.&lt;/p&gt;

&lt;p&gt;&lt;em&gt;List prices, on demand, before discounts. A one-year reservation trims the server and database lines by roughly a third, in exchange for paying for capacity you may never use. AWS's always-free allowances would cover the CDN and transfer lines at this size; they are shown at list price because they are the lines that grow with success.&lt;/em&gt;&lt;/p&gt;

&lt;h2&gt;
  
  
  What the bill does not show
&lt;/h2&gt;

&lt;p&gt;The monthly figure is the easy part. Every line in it is also a thing you have to build, and then keep.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Servers.&lt;/strong&gt; Choose an image, harden it, open the right ports and only those, install a runtime, write a deploy pipeline, put TLS in front of it, and then keep it patched for the rest of its life. Two servers means doing everything twice and keeping them identical. Autoscaling means an image pipeline on top of that.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;The database.&lt;/strong&gt; A subnet group, a parameter group, a security group, a backup window, a maintenance window, a failover test you actually run, connection pooling so a traffic spike does not exhaust it, and a plan for the day the disk fills up. Multi-AZ doubles the price, and it is the only version you would put a business on.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Files and the CDN.&lt;/strong&gt; A bucket with public access blocked, a policy that lets your servers write and nobody else, presigned URLs so users can upload without going through your servers, CORS rules, lifecycle rules. Then a CloudFront distribution with origin access control, cache policies, invalidations, a custom domain, and a certificate that, for CloudFront, must be issued in N. Virginia whatever region your files live in.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;The network.&lt;/strong&gt; A VPC, public and private subnets in two zones, route tables, an internet gateway, a NAT gateway, and IAM roles narrow enough to be safe and wide enough to work. None of this serves a single request. It is the price of admission.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;And then the months after.&lt;/strong&gt; Patches, certificate renewals, disk alarms, the 3 a.m. page, the database upgrade you have been putting off. If one developer spends even four hours a month keeping this running, that time alone costs more than the entire Skapi plan below, at any rate you would pay a developer in São Paulo, Buenos Aires or Bogotá.&lt;/p&gt;

&lt;h2&gt;
  
  
  Automated emails, compared
&lt;/h2&gt;

&lt;p&gt;Email is the line that looks free and is not. SES charges $0.10 per thousand messages, so 5,000 emails cost fifty cents in São Paulo, exactly as in Virginia. The cost is everything around that number.&lt;/p&gt;

&lt;p&gt;Before SES will send a real email you verify a domain, publish DKIM, SPF and DMARC records, and apply to leave the sandbox with a written description of how you handle bounces and complaints. Then you build that handling: notification topics, a suppression list, retries. Then the actual emails: a verification flow with a code that expires, a welcome email, an invitation, each one a template, a trigger and a queue so a burst of signups does not hit a rate limit. If you want to send a newsletter, add subscriber storage, a subscribe and unsubscribe flow that complies with the law in every country you send to, and a sending pipeline that paces itself.&lt;/p&gt;

&lt;p&gt;On Skapi that whole list is a settings page. Verification, welcome and invitation emails go out from your project's own sending address with no code. Newsletters have subscribers, subscription and unsubscribe links, and a send button. Standard includes 5,000 sends a month and Premium 50,000, and the deliverability plumbing is done once, for everyone, and kept warm.&lt;/p&gt;

&lt;h2&gt;
  
  
  The same backend on Skapi
&lt;/h2&gt;

&lt;p&gt;Same users, same data, same traffic, same city:&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2F7q2j9u5n5xmwwfe1s81a.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2F7q2j9u5n5xmwwfe1s81a.png" alt="Backend with Skapi" width="800" height="687"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;28 times less than the production stack. 13 times less than the one that goes down when you deploy.&lt;/strong&gt;&lt;/p&gt;

&lt;p&gt;Over a year, that is $6,345 against $228. The difference is not a discount. It is a different way of running a backend.&lt;/p&gt;

&lt;p&gt;Here is what each side gives you for the money, one line at a time. First the cheapest of each: the AWS bare minimum next to Skapi Standard.&lt;/p&gt;

&lt;div class="table-wrapper-paragraph"&gt;&lt;table&gt;
&lt;thead&gt;
&lt;tr&gt;
&lt;th&gt;&lt;/th&gt;
&lt;th&gt;AWS, bare minimum&lt;/th&gt;
&lt;th&gt;Skapi Standard&lt;/th&gt;
&lt;/tr&gt;
&lt;/thead&gt;
&lt;tbody&gt;
&lt;tr&gt;
&lt;td&gt;&lt;strong&gt;Price per month&lt;/strong&gt;&lt;/td&gt;
&lt;td&gt;&lt;strong&gt;$255.85&lt;/strong&gt;&lt;/td&gt;
&lt;td&gt;&lt;strong&gt;$19&lt;/strong&gt;&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Database&lt;/td&gt;
&lt;td&gt;RDS PostgreSQL with 2 vCPU and 4 GB of memory, one zone, 20 GB disk: $114.61&lt;/td&gt;
&lt;td&gt;3 GB, included&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;File storage&lt;/td&gt;
&lt;td&gt;S3, 120 GB: $6.68&lt;/td&gt;
&lt;td&gt;120 GB, included&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Web hosting&lt;/td&gt;
&lt;td&gt;Served from your one server, or from a bucket you configure&lt;/td&gt;
&lt;td&gt;Included, with a subdomain and HTTPS&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;CDN and bandwidth&lt;/td&gt;
&lt;td&gt;CloudFront 100 GB, 2 million requests, 30 GB of API traffic: $19.90&lt;/td&gt;
&lt;td&gt;100 GB, included&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;User accounts&lt;/td&gt;
&lt;td&gt;Cognito Lite, 20,000 active users: $55.00&lt;/td&gt;
&lt;td&gt;50,000 accounts, included&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;OpenID login&lt;/td&gt;
&lt;td&gt;Set up by you in Cognito&lt;/td&gt;
&lt;td&gt;Included&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Automated emails&lt;/td&gt;
&lt;td&gt;SES, 5,000 messages: $0.50. Templates built by you&lt;/td&gt;
&lt;td&gt;5,000 sends a month, included&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Newsletters&lt;/td&gt;
&lt;td&gt;Built by you: list, opt-out link, sending&lt;/td&gt;
&lt;td&gt;5,000 subscribers&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Realtime data&lt;/td&gt;
&lt;td&gt;Built by you, on the one server&lt;/td&gt;
&lt;td&gt;Included&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Push notifications&lt;/td&gt;
&lt;td&gt;Built by you&lt;/td&gt;
&lt;td&gt;Included&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Webhooks&lt;/td&gt;
&lt;td&gt;Endpoints you write and host&lt;/td&gt;
&lt;td&gt;Included&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;API secret keys&lt;/td&gt;
&lt;td&gt;Kept on your server&lt;/td&gt;
&lt;td&gt;Included&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;App server&lt;/td&gt;
&lt;td&gt;1 × t3.medium, 30 GB disk with snapshots: $55.66&lt;/td&gt;
&lt;td&gt;None to run&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Monitoring and DNS&lt;/td&gt;
&lt;td&gt;CloudWatch and Route 53: $3.50&lt;/td&gt;
&lt;td&gt;Usage meters in the dashboard&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Redundant&lt;/td&gt;
&lt;td&gt;No: one server, one zone&lt;/td&gt;
&lt;td&gt;Yes&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Backend deploys&lt;/td&gt;
&lt;td&gt;The site goes down each time&lt;/td&gt;
&lt;td&gt;None needed&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Scaling&lt;/td&gt;
&lt;td&gt;A bigger server, by hand&lt;/td&gt;
&lt;td&gt;Automatic&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Setup&lt;/td&gt;
&lt;td&gt;A day&lt;/td&gt;
&lt;td&gt;Minutes&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Upkeep&lt;/td&gt;
&lt;td&gt;Yours&lt;/td&gt;
&lt;td&gt;Ours&lt;/td&gt;
&lt;/tr&gt;
&lt;/tbody&gt;
&lt;/table&gt;&lt;/div&gt;

&lt;p&gt;And the full version of each: the AWS production stack next to Skapi Premium. To be generous to AWS, we added database overage to Premium until it matches the 20 GB database disk of the AWS stack, even though Premium already outdoes that stack everywhere else: five times the files, four times the bandwidth, ten times the emails.&lt;/p&gt;

&lt;div class="table-wrapper-paragraph"&gt;&lt;table&gt;
&lt;thead&gt;
&lt;tr&gt;
&lt;th&gt;&lt;/th&gt;
&lt;th&gt;AWS, production&lt;/th&gt;
&lt;th&gt;Skapi Premium&lt;/th&gt;
&lt;/tr&gt;
&lt;/thead&gt;
&lt;tbody&gt;
&lt;tr&gt;
&lt;td&gt;&lt;strong&gt;Price per month&lt;/strong&gt;&lt;/td&gt;
&lt;td&gt;&lt;strong&gt;$528.71&lt;/strong&gt;&lt;/td&gt;
&lt;td&gt;
&lt;strong&gt;$94&lt;/strong&gt; ($89 plan, $5 overage)&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Database&lt;/td&gt;
&lt;td&gt;RDS PostgreSQL with 2 vCPU and 4 GB of memory, two zones, 20 GB disk: $229.22&lt;/td&gt;
&lt;td&gt;20 GB: 10 GB included, 10 GB more at $0.50 per GB, $5.00&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;File storage&lt;/td&gt;
&lt;td&gt;S3, 120 GB: $6.68&lt;/td&gt;
&lt;td&gt;600 GB, included&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Web hosting&lt;/td&gt;
&lt;td&gt;Served from your servers, or from a bucket you configure&lt;/td&gt;
&lt;td&gt;Included, with a subdomain and HTTPS&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;CDN and bandwidth&lt;/td&gt;
&lt;td&gt;CloudFront 100 GB, 2 million requests, 30 GB of API traffic: $19.90&lt;/td&gt;
&lt;td&gt;400 GB, included&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;User accounts&lt;/td&gt;
&lt;td&gt;Cognito Lite, 20,000 active users: $55.00&lt;/td&gt;
&lt;td&gt;100,000 accounts, included&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;OpenID login&lt;/td&gt;
&lt;td&gt;Set up by you in Cognito&lt;/td&gt;
&lt;td&gt;Included&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Automated emails&lt;/td&gt;
&lt;td&gt;SES, 5,000 messages: $0.50. Templates built by you&lt;/td&gt;
&lt;td&gt;50,000 sends a month, included&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Newsletters&lt;/td&gt;
&lt;td&gt;Built by you: list, opt-out link, sending&lt;/td&gt;
&lt;td&gt;50,000 subscribers&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Realtime data&lt;/td&gt;
&lt;td&gt;Built by you, across two servers&lt;/td&gt;
&lt;td&gt;Included&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Push notifications&lt;/td&gt;
&lt;td&gt;Built by you&lt;/td&gt;
&lt;td&gt;Included&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Webhooks&lt;/td&gt;
&lt;td&gt;Endpoints you write and host&lt;/td&gt;
&lt;td&gt;Included&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;API secret keys&lt;/td&gt;
&lt;td&gt;Kept on your servers&lt;/td&gt;
&lt;td&gt;Included&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;App servers&lt;/td&gt;
&lt;td&gt;2 × t3.medium, 2 × 30 GB disk with snapshots: $111.31&lt;/td&gt;
&lt;td&gt;None to run&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Load balancer&lt;/td&gt;
&lt;td&gt;One capacity unit: $32.85&lt;/td&gt;
&lt;td&gt;None to run&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;NAT gateway&lt;/td&gt;
&lt;td&gt;For the private subnets: $69.75&lt;/td&gt;
&lt;td&gt;None to run&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Monitoring and DNS&lt;/td&gt;
&lt;td&gt;CloudWatch and Route 53: $3.50&lt;/td&gt;
&lt;td&gt;Usage meters in the dashboard&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Redundant&lt;/td&gt;
&lt;td&gt;Yes: two zones, set up and tested by you&lt;/td&gt;
&lt;td&gt;Yes&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Backend deploys&lt;/td&gt;
&lt;td&gt;Rolling, through a pipeline you build&lt;/td&gt;
&lt;td&gt;None needed&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Scaling&lt;/td&gt;
&lt;td&gt;Rules you write and tune&lt;/td&gt;
&lt;td&gt;Automatic&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Setup&lt;/td&gt;
&lt;td&gt;Days&lt;/td&gt;
&lt;td&gt;Minutes&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Upkeep&lt;/td&gt;
&lt;td&gt;Yours&lt;/td&gt;
&lt;td&gt;Ours&lt;/td&gt;
&lt;/tr&gt;
&lt;/tbody&gt;
&lt;/table&gt;&lt;/div&gt;

&lt;p&gt;That is $94 against $528.71. Even with the overage counted, Premium costs more than 5 times less.&lt;/p&gt;

&lt;p&gt;&lt;em&gt;The prices in each AWS column add up to the total at the top of it. The lines without a price are the ones the bill leaves out: they are paid in time instead. Both AWS columns are the bills from earlier in this article, sized for 120 GB of files and 20,000 active users. Sized for what Premium includes (600 GB of files, 400 GB of bandwidth, 50,000 emails, 40,000 active users) on the same servers, the AWS production bill comes to about $728 a month.&lt;/em&gt;&lt;/p&gt;

&lt;p&gt;Standard is $19 a month in São Paulo, the same as in Oregon, Paris, Mumbai, Singapore or Seoul. For that you get the database, file storage, CDN, web hosting, user accounts with OpenID login, automated emails and newsletters, realtime data, API secret keys and webhooks, with room for 50,000 user accounts, 3 GB of database, 120 GB of files and 100 GB of bandwidth. Premium is $89 for 100,000 accounts, 10 GB of database, 600 GB of files and 400 GB of bandwidth. A small project runs on Free for nothing at all.&lt;/p&gt;

&lt;p&gt;And all of that is before counting a single hour of anyone's time. The four hours a month from the section above cost more than the Skapi plan on their own.&lt;/p&gt;

&lt;h2&gt;
  
  
  How we got the price down
&lt;/h2&gt;

&lt;p&gt;We are not going to draw you our architecture, and we do not have secret hardware. What we have is a set of rules we did not break, and in São Paulo they matter more than anywhere else.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Nothing idles.&lt;/strong&gt; Look back at the first table. The steep premiums all sit on things billed by the hour whether or not anyone is using them. A traditional backend in São Paulo pays for 730 hours of servers, 730 hours of database and 730 hours of gateway every month, at 3 a.m. on a Sunday as much as at noon on payday. Every part of Skapi is metered by use: a request, a record read, a gigabyte stored, a message sent. When your users are asleep, so is your bill. That is what serverless means to us, and we applied it to every resource in the chain, not just the code.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Fixed costs are shared, not repeated.&lt;/strong&gt; The expensive lines above, the redundant database, the load balancer, the gateway, are fixed costs a single project pays alone. In Skapi those costs exist once per region, and every project in that region shares them. Your project's share of a fixed cost is close to zero, and that, more than any clever trick, is where the fraction comes from.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Bytes go to the cheapest place that can hold them.&lt;/strong&gt; Records live in a database that charges for what you store and read, not for a machine that waits. Anything large, files, uploads, hosted sites, record payloads over 32 KB, goes straight to object storage and is served from the CDN, so a download never touches a compute layer at all. Database storage is the most expensive byte in São Paulo, so we keep as few of them there as we can, and &lt;a href="https://dev.to/articles/2026-09-16"&gt;we made that change on every plan&lt;/a&gt;.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Identity and email are services, not servers.&lt;/strong&gt; User accounts live in a managed directory built for that job, with verification, password reset and multi-factor flows already there. Email is a queue, with delivery, bounces, complaints and sender reputation handled once for the whole platform. These are the parts that cost the same in every region, and they are the parts nobody should rebuild per project.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Scale is a consequence, not a plan.&lt;/strong&gt; Because every part is metered, a spike in traffic is a larger bill for that hour, not an outage, and a quiet month is a smaller one. There is no instance size to pick, no capacity to reserve, no failover to rehearse. The pieces that would go down are the pieces we do not run.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Security comes with the shape.&lt;/strong&gt; There are no servers in your project to patch, no SSH, no open ports. Every request is authenticated and authorized on its own, every component talks to the next with the narrowest permission that works, data is encrypted in transit and at rest, and private records can be &lt;a href="https://dev.to/articles/2026-09-07-02"&gt;encrypted in the browser&lt;/a&gt; under your user's own password. A project's data never leaves the region it was created in. For Latin American teams, that last line often matters as much as the price.&lt;/p&gt;

&lt;h2&gt;
  
  
  Why São Paulo was hard
&lt;/h2&gt;

&lt;p&gt;Here is the part we promised to explain. The rules above are how Skapi works everywhere. São Paulo tested every one of them.&lt;/p&gt;

&lt;p&gt;Serverless does not escape the regional premium. It meets it on a different bill. Every meter we rely on runs faster in São Paulo: storage per gigabyte, requests per million, bytes out per gigabyte. Some managed services cost more there, some arrived later than in other regions, and a few of the shared pieces that make a fixed cost disappear had to be planned again so they would still disappear at São Paulo prices. Keeping the plan price flat, $19 in São Paulo exactly as in Oregon, meant wasting nothing: fewer hops per request, caching wherever it pays for itself, moving bytes as little as possible, and checking that no component, anywhere in the chain, sits idle on your behalf.&lt;/p&gt;

&lt;p&gt;It was the hardest region we have opened, and we opened it on purpose. We want developers in Brazil and across Latin America to have affordable, efficient, solid access to backend infrastructure in their own region, at the same price as everyone else, without the tax the region has quietly charged for years. A student in Recife and a startup in Santiago should be able to run something real for the price of a few coffees a month. Now they can.&lt;/p&gt;

&lt;h2&gt;
  
  
  Where to start
&lt;/h2&gt;

&lt;p&gt;Create a project, choose São Paulo, and run the numbers against your own AWS bill. If the difference is not obvious in the first month, we would like to hear about it.&lt;/p&gt;

</description>
      <category>backend</category>
      <category>database</category>
      <category>aws</category>
      <category>cloud</category>
    </item>
    <item>
      <title>Olá, São Paulo: Skapi now runs in São Paulo, its first South American region</title>
      <dc:creator>Skapi</dc:creator>
      <pubDate>Wed, 30 Sep 2026 10:30:40 +0000</pubDate>
      <link>https://dev.to/skapi_api/ola-sao-paulo-skapi-now-runs-in-sao-paulo-its-first-south-american-region-42g7</link>
      <guid>https://dev.to/skapi_api/ola-sao-paulo-skapi-now-runs-in-sao-paulo-its-first-south-american-region-42g7</guid>
      <description>&lt;h2&gt;
  
  
  Olá, São Paulo
&lt;/h2&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fcdn.broadwayinc.com%2Fimg%2FFlag_Brazil.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fcdn.broadwayinc.com%2Fimg%2FFlag_Brazil.png" alt="Flag of Brazil" width="500" height="350"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;Skapi now runs in Brazil. São Paulo is our newest region, and it is the first place in South America where a project's database, its files and its user accounts all live on the same continent as the people using them.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fcdn.broadwayinc.com%2Fimg%2Fskapi-logo-cut-small.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fcdn.broadwayinc.com%2Fimg%2Fskapi-logo-cut-small.png" alt="Skapi" width="244" height="300"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;Olá, Brasil! Olá, América Latina! A partir de hoje, a Skapi roda em São Paulo. Se você cria aplicações para usuários no Brasil ou em qualquer canto da América Latina, eles não precisam mais cruzar meio mundo a cada requisição: o banco de dados, os arquivos e as contas do seu projeto ficam pertinho, e cada login, cada consulta e cada upload responde bem mais rápido. Um abraço, e bom código!&lt;/p&gt;

&lt;h2&gt;
  
  
  The first South American region
&lt;/h2&gt;

&lt;p&gt;Until now, the closest Skapi region to a user in Rio de Janeiro was in Paris, 9,167 km away on the other side of the Atlantic. A user in Lima was sent to Oregon, 7,720 km to the north. Every login, every query, every file upload made that journey and came back. It worked, and it worked well, but physics always took its cut.&lt;/p&gt;

&lt;p&gt;São Paulo changes the shape of that map. Here is what the same requests look like now:&lt;/p&gt;

&lt;div class="table-wrapper-paragraph"&gt;&lt;table&gt;
&lt;thead&gt;
&lt;tr&gt;
&lt;th&gt;Your user is in&lt;/th&gt;
&lt;th&gt;To São Paulo&lt;/th&gt;
&lt;th&gt;Nearest region before&lt;/th&gt;
&lt;/tr&gt;
&lt;/thead&gt;
&lt;tbody&gt;
&lt;tr&gt;
&lt;td&gt;Rio de Janeiro&lt;/td&gt;
&lt;td&gt;361 km&lt;/td&gt;
&lt;td&gt;9,167 km (Paris)&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Belo Horizonte&lt;/td&gt;
&lt;td&gt;491 km&lt;/td&gt;
&lt;td&gt;8,913 km (Paris)&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Porto Alegre&lt;/td&gt;
&lt;td&gt;852 km&lt;/td&gt;
&lt;td&gt;10,254 km (Paris)&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Brasília&lt;/td&gt;
&lt;td&gt;872 km&lt;/td&gt;
&lt;td&gt;8,725 km (Paris)&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Buenos Aires&lt;/td&gt;
&lt;td&gt;1,675 km&lt;/td&gt;
&lt;td&gt;10,852 km (Oregon)&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Santiago&lt;/td&gt;
&lt;td&gt;2,585 km&lt;/td&gt;
&lt;td&gt;10,099 km (Oregon)&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Lima&lt;/td&gt;
&lt;td&gt;3,454 km&lt;/td&gt;
&lt;td&gt;7,720 km (Oregon)&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Bogotá&lt;/td&gt;
&lt;td&gt;4,328 km&lt;/td&gt;
&lt;td&gt;6,339 km (Oregon)&lt;/td&gt;
&lt;/tr&gt;
&lt;/tbody&gt;
&lt;/table&gt;&lt;/div&gt;

&lt;p&gt;A user in Rio de Janeiro is now twenty-five times closer to their data. A user in Buenos Aires, six times closer. That difference is not a benchmark you have to squint at. It is the feeling of a page that loads before the spinner has a reason to appear, a form that saves the instant it is submitted, and a file that starts downloading rather than thinking about it first.&lt;/p&gt;

&lt;p&gt;From Colombia and Venezuela south, São Paulo is now the nearest region by a wide margin. Users in Mexico and most of Central America are still closest to Oregon, and Skapi keeps offering Oregon first there.&lt;/p&gt;

&lt;h2&gt;
  
  
  Where Skapi runs
&lt;/h2&gt;

&lt;p&gt;Six regions across North America, South America, Europe and Asia, placed so that most of the connected world has a Skapi region within a short hop.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Flhd1rbicahq7wp48psg1.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Flhd1rbicahq7wp48psg1.png" alt="Skapi covers 6 regions including LATAM" width="800" height="319"&gt;&lt;/a&gt;&lt;br&gt;
&lt;em&gt;Every dot is a full Skapi region running the complete stack.&lt;/em&gt;&lt;/p&gt;

&lt;div class="table-wrapper-paragraph"&gt;&lt;table&gt;
&lt;thead&gt;
&lt;tr&gt;
&lt;th&gt;Region&lt;/th&gt;
&lt;th&gt;Where it is&lt;/th&gt;
&lt;th&gt;Project ID starts with&lt;/th&gt;
&lt;/tr&gt;
&lt;/thead&gt;
&lt;tbody&gt;
&lt;tr&gt;
&lt;td&gt;us-west-2&lt;/td&gt;
&lt;td&gt;Oregon, United States&lt;/td&gt;
&lt;td&gt;&lt;code&gt;us72&lt;/code&gt;&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;sa-east-1&lt;/td&gt;
&lt;td&gt;&lt;strong&gt;São Paulo, Brazil&lt;/strong&gt;&lt;/td&gt;
&lt;td&gt;&lt;code&gt;sa31&lt;/code&gt;&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;eu-west-3&lt;/td&gt;
&lt;td&gt;Paris, France&lt;/td&gt;
&lt;td&gt;&lt;code&gt;eu73&lt;/code&gt;&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;ap-south-1&lt;/td&gt;
&lt;td&gt;Mumbai, India&lt;/td&gt;
&lt;td&gt;&lt;code&gt;ap51&lt;/code&gt;&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;ap-southeast-1&lt;/td&gt;
&lt;td&gt;Singapore&lt;/td&gt;
&lt;td&gt;&lt;code&gt;ap41&lt;/code&gt;&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;ap-northeast-2&lt;/td&gt;
&lt;td&gt;Seoul, South Korea&lt;/td&gt;
&lt;td&gt;&lt;code&gt;ap22&lt;/code&gt;&lt;/td&gt;
&lt;/tr&gt;
&lt;/tbody&gt;
&lt;/table&gt;&lt;/div&gt;

&lt;p&gt;Read that list from west to east and the spacing is deliberate. Oregon anchors North America, and São Paulo now does the same for South America. Paris covers Europe, North Africa and the Middle East. Mumbai, Singapore and Seoul carry the largest concentration of internet users on the planet between them.&lt;/p&gt;

&lt;p&gt;Every one of those regions runs the same Skapi. The same SDK call, the same API, the same authentication, the same record model. There is no South American edition with fewer features and no waiting for a region to catch up. A region is either serving your project completely or it is not serving it at all.&lt;/p&gt;

&lt;h2&gt;
  
  
  Now you choose where your project lives
&lt;/h2&gt;

&lt;p&gt;When you create a project, you pick the region.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto/https%3A%2F%2Fcdn.broadwayinc.com%2Farticles%2Fimg%2Fcreate-project-region.webp" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto/https%3A%2F%2Fcdn.broadwayinc.com%2Farticles%2Fimg%2Fcreate-project-region.webp" alt="The Create a new project dialog: a project name, an optional description, and a Region field reading Sao Paulo, Brazil (closest to you), with a note that the project's data is stored in the region you pick and cannot be moved later" width="800" height="650"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;&lt;em&gt;The new-project dialog. The closest region is offered first, every region is in the list, and the note under it says the one thing to know before you continue.&lt;/em&gt;&lt;/p&gt;

&lt;p&gt;Skapi offers you the closest one first, worked out from where you are, so the fast choice is also the default choice. But the list is right there, and the right answer is often not your own city. If you are a studio in Lisbon with an audience in Rio, put the project in São Paulo. If you are three founders in São Paulo selling to Madrid, put it in Paris. Your users are the ones making the round trip thousands of times a day. You make it twice.&lt;/p&gt;

&lt;p&gt;Pick deliberately, because a project stays where you put it. The region is baked into the first four characters of the project ID, which is why a São Paulo project announces itself as &lt;code&gt;sa31...&lt;/code&gt; for the rest of its life. It is a permanent decision in the same way a good foundation is permanent.&lt;/p&gt;

&lt;h2&gt;
  
  
  Data that stays where you put it
&lt;/h2&gt;

&lt;p&gt;There is a second reason to care about a region. A project created in São Paulo keeps its database records, its uploaded files, its hosted site and its user accounts in Brazil. Not routed through it. Stored in it.&lt;/p&gt;

&lt;p&gt;For teams whose customers ask where their data is kept, and for anyone working under the LGPD, Brazil's data protection law, that single sentence is usually the one that unblocks the project.&lt;/p&gt;

&lt;h2&gt;
  
  
  Why São Paulo means so much to us
&lt;/h2&gt;

&lt;p&gt;A new region is always satisfying to open, but this one means more to us than most, for two reasons.&lt;/p&gt;

&lt;p&gt;The first is reach. South America is home to hundreds of millions of people who live on their phones, and until now every one of them was a long way from any Skapi project. That distance showed up in your business as slower pages, abandoned forms and users who quietly gave up. With São Paulo, your product can reach them properly. It loads faster, it saves faster, and it feels like it was built for them, because now it runs right next to them.&lt;/p&gt;

&lt;p&gt;The second is cost. Running a traditional server backend on AWS in South America is expensive, noticeably more so than the same servers in North America or Europe, and that cost alone has kept many teams from serving the region well, or at all. We made sure a Skapi project in São Paulo runs at a fraction of that price. Making that possible was a real challenge for us, and it took a lot more than flipping a switch. That story deserves its own space, so we will cover it in the next article.&lt;/p&gt;

&lt;p&gt;So: create a project, choose São Paulo, and let your Brazilian and Latin American users feel what their applications were capable of all along.&lt;/p&gt;

&lt;p&gt;Bem-vindos a bordo. Bom desenvolvimento, e até breve.&lt;/p&gt;

</description>
      <category>security</category>
      <category>database</category>
      <category>aws</category>
      <category>cloud</category>
    </item>
    <item>
      <title>Skapi 2.0 Is Out: Private Data Encryption, API Streaming, and Safer Backend Requests</title>
      <dc:creator>Skapi</dc:creator>
      <pubDate>Mon, 07 Sep 2026 09:54:02 +0000</pubDate>
      <link>https://dev.to/skapi_api/skapi-20-is-out-private-data-encryption-api-streaming-and-safer-backend-requests-156b</link>
      <guid>https://dev.to/skapi_api/skapi-20-is-out-private-data-encryption-api-streaming-and-safer-backend-requests-156b</guid>
      <description>&lt;h2&gt;
  
  
  Skapi 2.0 is out
&lt;/h2&gt;

&lt;p&gt;The stable tag has moved: &lt;code&gt;npm i skapi-js&lt;/code&gt; installs 2.0 today. Private records can now be encrypted in the browser, under your user's own password.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto/https%3A%2F%2Fwww.bunnyquery.com%2Fskapi-square-128.webp" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto/https%3A%2F%2Fwww.bunnyquery.com%2Fskapi-square-128.webp" alt="Skapi" width="128" height="128"&gt;&lt;/a&gt;&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight plaintext"&gt;&lt;code&gt;npm i skapi-js
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;In August that command still installed 1.8.3, and we said nothing you were running would change until you asked for it. This is you asking for it. If you are not ready, &lt;code&gt;npm i skapi-js@1.8.3&lt;/code&gt; keeps you exactly where you are.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://www.bunnyquery.com/articles/2026-08-12" rel="noopener noreferrer"&gt;The release candidate post&lt;/a&gt; covered the naming, the polling queues and the session storage work. This one is about what landed on top of them.&lt;/p&gt;

&lt;h2&gt;
  
  
  Private records can be encrypted in the browser now
&lt;/h2&gt;

&lt;p&gt;&lt;code&gt;new Skapi("&amp;lt;Project ID&amp;gt;", { encryption: true })&lt;/code&gt; encrypts the &lt;code&gt;data&lt;/code&gt; of every record you write to &lt;code&gt;access_group: 'private'&lt;/code&gt;, and the contents of the files attached to it, before any of it leaves the page. The key is derived from the user's own password and never leaves their device.&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight javascript"&gt;&lt;code&gt;&lt;span class="kd"&gt;const&lt;/span&gt; &lt;span class="nx"&gt;skapi&lt;/span&gt; &lt;span class="o"&gt;=&lt;/span&gt; &lt;span class="k"&gt;new&lt;/span&gt; &lt;span class="nc"&gt;Skapi&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="dl"&gt;"&lt;/span&gt;&lt;span class="s2"&gt;&amp;lt;Project ID&amp;gt;&lt;/span&gt;&lt;span class="dl"&gt;"&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="p"&gt;{&lt;/span&gt;
    &lt;span class="na"&gt;encryption&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="p"&gt;{&lt;/span&gt; &lt;span class="na"&gt;iterations&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="mi"&gt;600000&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="na"&gt;minPasswordLength&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="mi"&gt;12&lt;/span&gt; &lt;span class="p"&gt;}&lt;/span&gt;
&lt;span class="p"&gt;});&lt;/span&gt;

&lt;span class="k"&gt;await&lt;/span&gt; &lt;span class="nx"&gt;skapi&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;postRecord&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;
    &lt;span class="p"&gt;{&lt;/span&gt; &lt;span class="na"&gt;diagnosis&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="s1"&gt;confidential&lt;/span&gt;&lt;span class="dl"&gt;'&lt;/span&gt; &lt;span class="p"&gt;},&lt;/span&gt;
    &lt;span class="p"&gt;{&lt;/span&gt; &lt;span class="na"&gt;table&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="p"&gt;{&lt;/span&gt; &lt;span class="na"&gt;name&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="s1"&gt;notes&lt;/span&gt;&lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="na"&gt;access_group&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="s1"&gt;private&lt;/span&gt;&lt;span class="dl"&gt;'&lt;/span&gt; &lt;span class="p"&gt;}&lt;/span&gt; &lt;span class="p"&gt;}&lt;/span&gt;
&lt;span class="p"&gt;);&lt;/span&gt;
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;The calls do not change. &lt;code&gt;postRecord()&lt;/code&gt; seals, &lt;code&gt;getRecords()&lt;/code&gt; opens, and there is no key for your end user to manage. Sharing is the call it always was: &lt;code&gt;grantPrivateRecordAccess()&lt;/code&gt; wraps that record's data key to the recipient's published public key, so the owner does not have to be online when the grantee reads, and &lt;code&gt;removePrivateRecordAccess()&lt;/code&gt; rolls the data key, which makes revocation forward only.&lt;/p&gt;

&lt;p&gt;Skapi has always decided who may read a record, on the server, against a signed identity the browser cannot forge. What is new is that the bytes stop being readable by the people running the database. In the words our documentation uses: a provider who dumps the database, the storage bucket and every request log cannot recover that record's &lt;code&gt;data&lt;/code&gt; without a successful offline guess against the user's password.&lt;/p&gt;

&lt;h2&gt;
  
  
  An API response can render while it is still being written
&lt;/h2&gt;

&lt;p&gt;&lt;code&gt;clientSecretRequest()&lt;/code&gt; now takes &lt;code&gt;stream: true&lt;/code&gt; and &lt;code&gt;onStream(chunk, seq, via)&lt;/code&gt;. The server reads the destination's response incrementally and relays the raw bytes as they arrive, so your page renders an answer while the destination is still writing it. Add &lt;code&gt;realtime: true&lt;/code&gt; and each piece also comes over skapi's websocket the moment it is relayed, instead of on the next poll tick.&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight javascript"&gt;&lt;code&gt;&lt;span class="nx"&gt;skapi&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;clientSecretRequest&lt;/span&gt;&lt;span class="p"&gt;({&lt;/span&gt;
    &lt;span class="na"&gt;clientSecretName&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="s1"&gt;openai&lt;/span&gt;&lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt;
    &lt;span class="na"&gt;url&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="s1"&gt;https://api.openai.com/v1/chat/completions&lt;/span&gt;&lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt;
    &lt;span class="na"&gt;method&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="s1"&gt;POST&lt;/span&gt;&lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt;
    &lt;span class="na"&gt;headers&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="p"&gt;{&lt;/span&gt; &lt;span class="na"&gt;Authorization&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="s1"&gt;Bearer $CLIENT_SECRET&lt;/span&gt;&lt;span class="dl"&gt;'&lt;/span&gt; &lt;span class="p"&gt;},&lt;/span&gt;
    &lt;span class="na"&gt;data&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="p"&gt;{&lt;/span&gt; &lt;span class="na"&gt;model&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="s1"&gt;gpt-4.1&lt;/span&gt;&lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="na"&gt;stream&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="kc"&gt;true&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="nx"&gt;messages&lt;/span&gt; &lt;span class="p"&gt;},&lt;/span&gt; &lt;span class="c1"&gt;// the destination's own flag&lt;/span&gt;
    &lt;span class="na"&gt;stream&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="kc"&gt;true&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt;    &lt;span class="c1"&gt;// skapi's flag: relay the answer as it arrives&lt;/span&gt;
    &lt;span class="na"&gt;realtime&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="kc"&gt;true&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt;
    &lt;span class="na"&gt;poll&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="mi"&gt;1000&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt;
    &lt;span class="na"&gt;onStream&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="nx"&gt;chunk&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="nx"&gt;seq&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="nx"&gt;via&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt; &lt;span class="o"&gt;=&amp;gt;&lt;/span&gt; &lt;span class="nf"&gt;readEvents&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="nx"&gt;chunk&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;
&lt;span class="p"&gt;});&lt;/span&gt;
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;Your key still never reaches the browser: &lt;code&gt;$CLIENT_SECRET&lt;/code&gt; is filled in on the server as the call goes out. Skapi parses nothing, so the framing and the buffering of a chunk that ended mid frame are yours, which is why this works against any destination on day one rather than only the ones we know about. The websocket is an accelerator and never the source of truth: a dropped socket is not an error, and the read continues at poll speed.&lt;/p&gt;

&lt;p&gt;One trap. &lt;code&gt;stream&lt;/code&gt; says how skapi reads the response; asking the destination to produce one incrementally is a separate flag in your own request body. Skapi inspects neither, so setting only one of the two goes wrong quietly, with no error from either side.&lt;/p&gt;

&lt;h2&gt;
  
  
  forwardRequest is the method we promised and never named
&lt;/h2&gt;

&lt;p&gt;The release candidate post said 2.0 had one genuinely new method, then no section ever said which. It is &lt;code&gt;forwardRequest()&lt;/code&gt;.&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight javascript"&gt;&lt;code&gt;&lt;span class="nx"&gt;skapi&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;forwardRequest&lt;/span&gt;&lt;span class="p"&gt;({&lt;/span&gt; &lt;span class="na"&gt;question&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="s1"&gt;How did last quarter go?&lt;/span&gt;&lt;span class="dl"&gt;'&lt;/span&gt; &lt;span class="p"&gt;},&lt;/span&gt; &lt;span class="p"&gt;{&lt;/span&gt;
    &lt;span class="na"&gt;url&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="s1"&gt;https://api.yourbackend.com/chat&lt;/span&gt;&lt;span class="dl"&gt;'&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt;
    &lt;span class="na"&gt;onStream&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="nx"&gt;chunk&lt;/span&gt; &lt;span class="o"&gt;=&amp;gt;&lt;/span&gt; &lt;span class="nx"&gt;output&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nx"&gt;textContent&lt;/span&gt; &lt;span class="o"&gt;+=&lt;/span&gt; &lt;span class="nx"&gt;chunk&lt;/span&gt;
&lt;span class="p"&gt;});&lt;/span&gt;
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;It relays a request to your own backend from skapi's servers and streams the response back, with your project's API key attached server side where the browser cannot read it. Your backend also receives &lt;code&gt;x-skapi-user&lt;/code&gt;, taken from the verified session, and a request that tries to set an &lt;code&gt;x-skapi-&lt;/code&gt; header itself is rejected, so that attribution is safe to trust. The first argument can be a submit event, a form element, a &lt;code&gt;FormData&lt;/code&gt; or a plain object, and a form goes as &lt;code&gt;multipart/form-data&lt;/code&gt; with its files.&lt;/p&gt;

&lt;p&gt;Your backend has to resolve to a public address, so private ranges, loopback and cloud metadata addresses are refused. The destination instructions ride in a header capped at 4096 characters measured after escaping, where each non-ASCII character costs six, so put anything long in the body. And there is no re-attach call for this one: the response is yours once, as it arrives.&lt;/p&gt;

&lt;h2&gt;
  
  
  Three things to know before you upgrade
&lt;/h2&gt;

&lt;p&gt;&lt;strong&gt;Your &lt;code&gt;data&lt;/code&gt; is stored as JSON text, and the SDK is the only decoder.&lt;/strong&gt; The API encodes it on every write regardless of which client wrote it, so an older SDK reading one of these records gets the raw encoded string instead of your object, quietly, with no error. The test is one line: a &lt;code&gt;data&lt;/code&gt; you expected as an object that reads back as a string beginning &lt;code&gt;!J%&lt;/code&gt; is this. If a project has more than one client on it, a mobile build, a Node worker, a REST caller, upgrade them together.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Two changes ship with the API rather than the SDK, so they are already true for you.&lt;/strong&gt; Moving a record into or out of &lt;code&gt;'private'&lt;/code&gt; clears every private access grant on it, encryption or not, so grant those users again after the move. And only a record's owner may cross that line: a master account still reads, edits and deletes anything in its project, but it can no longer move somebody else's record between &lt;code&gt;'private'&lt;/code&gt; and a numbered group, because with encryption in play that would destroy the record while reporting success.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Four small changes arrive whether you ask or not.&lt;/strong&gt; &lt;code&gt;skapi.util.decodeServiceId()&lt;/code&gt; is &lt;code&gt;decodeProjectId()&lt;/code&gt;, with no alias. &lt;code&gt;table: 'name'&lt;/code&gt; on an update no longer moves that record to the public access group, so pass &lt;code&gt;table: { name: 'name', access_group: 0 }&lt;/code&gt; if you wanted the move. &lt;code&gt;deleteRecords()&lt;/code&gt; returns normalized records, so read &lt;code&gt;record_id&lt;/code&gt; and &lt;code&gt;table.name&lt;/code&gt; rather than the old short keys. And &lt;code&gt;data: null&lt;/code&gt; stores a null instead of removing the field.&lt;/p&gt;

&lt;p&gt;Everything else waits to be asked for, including the new &lt;code&gt;require_login&lt;/code&gt; project setting, read from &lt;code&gt;getConnectionInfo()&lt;/code&gt; as &lt;code&gt;conf.require_login&lt;/code&gt;. It makes the SDK refuse record reads from a signed out session: a guard rail against a page leaking your public records by accident, not an access control, since the API still serves &lt;code&gt;access_group: 0&lt;/code&gt; to anyone who asks.&lt;/p&gt;

&lt;p&gt;One correction while you have &lt;a href="https://docs.skapi.com/versionlog/versions.html" rel="noopener noreferrer"&gt;the changelog&lt;/a&gt; open. &lt;code&gt;condition: '&amp;lt;='&lt;/code&gt; on a string index value as an "ends with" search is a 1.8.0 change, not a 2.0 one, and the API serves it, so it already works from whatever version you are running today.&lt;/p&gt;

&lt;h2&gt;
  
  
  The bugs we could not have found on our own
&lt;/h2&gt;

&lt;p&gt;None of the fixes in that data section came from reading our own code. A string containing &lt;code&gt;*add&lt;/code&gt; destroying a record is not something a test suite finds: somebody has to be storing a sentence a real person wrote, in a real app, before the storage format ever gets the chance to misread it.&lt;/p&gt;

&lt;p&gt;So thank you to everyone who put a project that mattered to them on a release candidate, kept &lt;code&gt;skapi-js@rc&lt;/code&gt; installed while it settled, and wrote in with the boring exact detail: the record id, the payload, the call that came back wrong. That is the part of 2.0 we could not have written ourselves.&lt;/p&gt;

&lt;p&gt;Encryption is the first place where Skapi holds something it genuinely cannot read, and widening that is where the next stretch goes. &lt;code&gt;data&lt;/code&gt; is sealed and &lt;code&gt;index.value&lt;/code&gt; is not, because a range read over ordered keys is what keeps a schemaless query cheap. Protecting what you query on without losing the ability to query it is a hard question, and we are not going to answer it badly.&lt;/p&gt;

&lt;p&gt;There is one more thing your backend does not have now. A copy of your users' data it can read.&lt;/p&gt;

&lt;p&gt;Go build with Skapi, and let your users keep their own keys.&lt;/p&gt;

</description>
      <category>programming</category>
      <category>productivity</category>
      <category>api</category>
      <category>backend</category>
    </item>
    <item>
      <title>The Fastest Way to Deploy a Site in 2026</title>
      <dc:creator>Skapi</dc:creator>
      <pubDate>Thu, 11 Dec 2025 08:01:39 +0000</pubDate>
      <link>https://dev.to/skapi_api/the-fastest-way-to-deploy-a-site-in-2026-a10</link>
      <guid>https://dev.to/skapi_api/the-fastest-way-to-deploy-a-site-in-2026-a10</guid>
      <description>&lt;blockquote&gt;
&lt;p&gt;“Most developers don’t need a backend. They need a fast way to launch.”&lt;/p&gt;
&lt;/blockquote&gt;

&lt;p&gt;By the end of 2025, the question many developers ask is no longer how to design a website or structure a backend. The question has become much simpler and far more urgent: &lt;strong&gt;What is the fastest way to deploy a site?&lt;/strong&gt;&lt;/p&gt;

&lt;p&gt;Time to launch has become a key competitive factor for indie developers, small teams, and even established engineers who want to move faster without getting caught in configuration loops and infrastructure overhead.&lt;/p&gt;

&lt;p&gt;This article explores that question through the lens of Alex Lee, an early-career developer who has already grown tired of unnecessary complexity, juggling tools, and backend setup. His experience reveals a shift happening in the industry, and why platforms like &lt;a href="//www.skapi.com"&gt;Skapi&lt;/a&gt; are quickly becoming the fastest and most reliable way to deploy a site with zero configuration and no backend work.&lt;/p&gt;

&lt;h2&gt;
  
  
  The Development Pain Point Alex Can’t Ignore:
&lt;/h2&gt;

&lt;p&gt;Alex Lee, a developer with two years of experience and a full-time job, decided to work on side projects he genuinely enjoyed. He was excited to get started, but quickly ran into problems he never expected.&lt;/p&gt;

&lt;p&gt;His idea was clear, the frontend code was ready, and the user flow was straightforward. He wasn’t building a large-scale SaaS platform or a complex distributed application. He just wanted to deploy a clean, static site – a ticket pre-sale platform for his favorite music band – with a few simple functional components: email handling, a form submission endpoint, and some basic state storage.&lt;/p&gt;

&lt;p&gt;Sounds simple, right? Looks even simpler:&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.amazonaws.com%2Fuploads%2Farticles%2Fih73wk2wrd9e57qkpqu8.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.amazonaws.com%2Fuploads%2Farticles%2Fih73wk2wrd9e57qkpqu8.png" alt="music group pre-sale platform website" width="800" height="370"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;Yet deploying that site turned into a cycle of dashboards, configs, environment variables, SSL settings, file structures, and DNS delays. What should have taken minutes routinely took hours. For an early-career developer, the complexity felt disproportionate to the goal.&lt;/p&gt;

&lt;p&gt;This frustration is increasingly common, and it reveals the core of a major trend: &lt;strong&gt;developers want the fastest possible way to deploy a site without touching backend infrastructure at all&lt;/strong&gt;.&lt;/p&gt;

&lt;h2&gt;
  
  
  The Search for the Fastest Way to Deploy a Site
&lt;/h2&gt;

&lt;p&gt;In exploring options, Alex tried several of the popular hosting and deployment tools that have defined the industry over the past few years. Each one promised simplicity. Each one advertised speed. Yet most of them came with underlying complexity that slowed down the process:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;build configurations&lt;/li&gt;
&lt;li&gt;routing files&lt;/li&gt;
&lt;li&gt;backend functions that required separation&lt;/li&gt;
&lt;li&gt;authentication that needed setup&lt;/li&gt;
&lt;li&gt;email APIs requiring external providers&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;Too much for a developer whose project did not require this level of machinery.&lt;/p&gt;

&lt;p&gt;Alex realized the actual question was not &lt;em&gt;What hosting tool should I use?&lt;/em&gt; but rather:&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Why does deploying a small, functional site still require thinking about backend infrastructure at all?&lt;/strong&gt;&lt;/p&gt;

&lt;p&gt;And this is where Alex found &lt;strong&gt;Skapi&lt;/strong&gt;.&lt;/p&gt;

&lt;h2&gt;
  
  
  A New Approach: Zero-Backend Deployment
&lt;/h2&gt;

&lt;p&gt;Skapi stood out immediately because it answered the one question that mattered:&lt;br&gt;
&lt;strong&gt;How fast can a developer deploy a site without touching backend configuration?&lt;/strong&gt;&lt;/p&gt;

&lt;p&gt;Alex discovered that Skapi’s architecture does something different from traditional hosting platforms:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;It hosts static sites instantly through a simple file-drop interface. index.html becomes your root entry point. Files are served instantly through global CDN.&lt;/li&gt;
&lt;li&gt;It removes the backend setup entirely.&lt;/li&gt;
&lt;li&gt;It provides automatic email handling and newsletter flows.&lt;/li&gt;
&lt;li&gt;It supports authentication without configuration.&lt;/li&gt;
&lt;li&gt;It includes built-in storage and database-like record handling.&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.amazonaws.com%2Fuploads%2Farticles%2F77dcx6m1dqjfx5smhjlq.jpg" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.amazonaws.com%2Fuploads%2Farticles%2F77dcx6m1dqjfx5smhjlq.jpg" alt="easy no-setup backend" width="800" height="392"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;In other words, Skapi collapses multiple layers of development work into a single environment that requires no manual setup. For a developer trying to deploy a side-project, it represented exactly what the industry has been moving toward: &lt;strong&gt;the fastest way to deploy a site with AI-assisted backend logic and zero configuration.&lt;/strong&gt;&lt;/p&gt;

&lt;blockquote&gt;
&lt;p&gt;“Skapi let me deploy my project without touching any infrastructure at all.”&lt;/p&gt;
&lt;/blockquote&gt;

&lt;h2&gt;
  
  
  Deploying the Side-Project: A Faster Workflow
&lt;/h2&gt;

&lt;p&gt;Alex’s deployment process shifted dramatically once he tried Skapi. &lt;/p&gt;

&lt;p&gt;The steps became straightforward:&lt;/p&gt;

&lt;ol&gt;
&lt;li&gt;Prepare the frontend folder.&lt;/li&gt;
&lt;li&gt;Drag files into Skapi’s File Hosting.&lt;/li&gt;
&lt;li&gt;The site is live instantly.&lt;/li&gt;
&lt;/ol&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.amazonaws.com%2Fuploads%2Farticles%2Fwk8g325injgm7xqwi0nm.jpg" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.amazonaws.com%2Fuploads%2Farticles%2Fwk8g325injgm7xqwi0nm.jpg" alt="instant web hosting in 2 minutes" width="800" height="395"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;There were no DNS tasks, no build steps, and no backend endpoints to configure. Alex was able to add a form submission flow with Skapi’s built-in API handling. He integrated email automation without any third-party providers. He added basic data storage without provisioning a database.&lt;/p&gt;

&lt;p&gt;This workflow matched exactly what many developers search for. When someone types &lt;strong&gt;the fastest way to deploy a site&lt;/strong&gt; or &lt;strong&gt;how to deploy a website quickly&lt;/strong&gt;, they’re not looking for server diagrams or configuration tutorials. They want a process that lets them build and deploy without obstacles.&lt;/p&gt;

&lt;p&gt;Skapi delivered that experience by treating backend complexity as optional rather than required.&lt;/p&gt;

&lt;p&gt;The industry is changing. Backends will always matter for large-scale systems, but most modern projects: side-project apps, landing pages, prototypes, microtools, and personal products do not require full backend environments. &lt;/p&gt;

&lt;p&gt;This gap between actual requirements and traditional infrastructure has opened the door for tools like Skapi to reshape deployment workflows entirely.&lt;/p&gt;

&lt;h2&gt;
  
  
  Happy Alex’s Conclusion
&lt;/h2&gt;

&lt;p&gt;Alex’s search for the fastest way to deploy a site ended not with a new tool to configure, but with a different approach to development itself. Skapi provided a clear path where the friction of backend setup disappeared, leaving space for building, refining, and launching a real side-project in a fraction of the time.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;The fastest way to deploy a site in 2026 is to use tools that eliminate backend work entirely.&lt;/strong&gt; For anyone searching for the quickest and most effective deployment workflow, the answer is becoming clearer each year: choose the tools that let you build without barriers. &lt;/p&gt;

&lt;p&gt;In that landscape, Skapi is quickly emerging as one of the strongest options.&lt;/p&gt;

&lt;p&gt;Follow us on socials: &lt;a href="https://x.com/skapi_api" rel="noopener noreferrer"&gt;X&lt;/a&gt;, &lt;a href="https://www.youtube.com/@skapi_official" rel="noopener noreferrer"&gt;YouTube&lt;/a&gt;, &lt;a href="https://www.instagram.com/skapi_api" rel="noopener noreferrer"&gt;Instagram&lt;/a&gt;, &lt;a href="https://www.linkedin.com/company/skapi-backend-api/" rel="noopener noreferrer"&gt;LinkedIn&lt;/a&gt;. More tutorials, code drops, and dev hacks are coming soon.&lt;/p&gt;

</description>
      <category>productivity</category>
      <category>automation</category>
      <category>api</category>
      <category>webdev</category>
    </item>
  </channel>
</rss>
