<?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: Aplite</title>
    <description>The latest articles on DEV Community by Aplite (@aplite).</description>
    <link>https://dev.to/aplite</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%2F4168153%2Fa7324411-ac17-4a14-ba7f-e5c13325ff99.png</url>
      <title>DEV Community: Aplite</title>
      <link>https://dev.to/aplite</link>
    </image>
    <atom:link rel="self" type="application/rss+xml" href="https://dev.to/feed/aplite"/>
    <language>en</language>
    <item>
      <title>7 ways to host OpenClaw for free in 2026, compared</title>
      <dc:creator>Aplite</dc:creator>
      <pubDate>Wed, 21 Oct 2026 03:38:25 +0000</pubDate>
      <link>https://dev.to/aplite/7-ways-to-host-openclaw-for-free-in-2026-compared-11m0</link>
      <guid>https://dev.to/aplite/7-ways-to-host-openclaw-for-free-in-2026-compared-11m0</guid>
      <description>&lt;p&gt;&lt;em&gt;I work on &lt;a href="https://aplite.ai/docs/openclaw?utm_source=devto&amp;amp;utm_medium=article&amp;amp;utm_campaign=post&amp;amp;utm_content=free-7-ways" rel="noopener noreferrer"&gt;Aplite&lt;/a&gt;, a managed OpenClaw host. It is one of the seven options below. The terms of the other six come from their own pages, read on 19 October 2026, and the links go there.&lt;/em&gt;&lt;/p&gt;

&lt;p&gt;"Free OpenClaw hosting" means four different things depending on who is selling it: a machine you already own, a cloud free tier where you install everything yourself, a managed plan with no end date, or free hosting with the tokens billed from the first reply. Two bills decide whether $0 is really $0. An agent that works around the clock needs a machine that stays on, and it needs model tokens for every reply. Each of the seven ways below covers one bill, both, or one of them for a while.&lt;/p&gt;

&lt;h2&gt;
  
  
  The seven, in one table
&lt;/h2&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;Way&lt;/th&gt;
&lt;th&gt;What is free&lt;/th&gt;
&lt;th&gt;Card at sign-up&lt;/th&gt;
&lt;th&gt;Your API key&lt;/th&gt;
&lt;th&gt;Free ends&lt;/th&gt;
&lt;/tr&gt;
&lt;/thead&gt;
&lt;tbody&gt;
&lt;tr&gt;
&lt;td&gt;1&lt;/td&gt;
&lt;td&gt;Your own machine, local model&lt;/td&gt;
&lt;td&gt;everything&lt;/td&gt;
&lt;td&gt;no&lt;/td&gt;
&lt;td&gt;no&lt;/td&gt;
&lt;td&gt;never&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;2&lt;/td&gt;
&lt;td&gt;Oracle Cloud Always Free&lt;/td&gt;
&lt;td&gt;a VM: 2 Arm OCPUs, 12 GB, or two 1 GB micro VMs&lt;/td&gt;
&lt;td&gt;yes, plus a phone number&lt;/td&gt;
&lt;td&gt;yes, or a local model&lt;/td&gt;
&lt;td&gt;never; idle VMs can be reclaimed&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;3&lt;/td&gt;
&lt;td&gt;Google Cloud free tier&lt;/td&gt;
&lt;td&gt;one e2-micro, 1 GB, 30 GB disk&lt;/td&gt;
&lt;td&gt;yes&lt;/td&gt;
&lt;td&gt;yes&lt;/td&gt;
&lt;td&gt;never&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;4&lt;/td&gt;
&lt;td&gt;AWS Free Tier&lt;/td&gt;
&lt;td&gt;$100 in credits, up to $200&lt;/td&gt;
&lt;td&gt;the page does not say&lt;/td&gt;
&lt;td&gt;yes&lt;/td&gt;
&lt;td&gt;6 months, or when the credits run out&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;5&lt;/td&gt;
&lt;td&gt;Railway, official template&lt;/td&gt;
&lt;td&gt;a $5 credit&lt;/td&gt;
&lt;td&gt;no&lt;/td&gt;
&lt;td&gt;yes&lt;/td&gt;
&lt;td&gt;30 days&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;6&lt;/td&gt;
&lt;td&gt;A managed free plan: Aplite, Agent37&lt;/td&gt;
&lt;td&gt;hosting, one agent, a starter credit&lt;/td&gt;
&lt;td&gt;no&lt;/td&gt;
&lt;td&gt;no on Aplite; on Agent37 after $0.20 of the default model&lt;/td&gt;
&lt;td&gt;never; both sleep between uses&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;7&lt;/td&gt;
&lt;td&gt;Free hosting, metered tokens: Runware, Ampere&lt;/td&gt;
&lt;td&gt;the instance&lt;/td&gt;
&lt;td&gt;no&lt;/td&gt;
&lt;td&gt;no&lt;/td&gt;
&lt;td&gt;hosting never; tokens from the first reply&lt;/td&gt;
&lt;/tr&gt;
&lt;/tbody&gt;
&lt;/table&gt;&lt;/div&gt;

&lt;h2&gt;
  
  
  1. Your own machine, with a local model
&lt;/h2&gt;

&lt;p&gt;The only way where neither bill exists. Install &lt;a href="https://ollama.com/download" rel="noopener noreferrer"&gt;Ollama&lt;/a&gt;, pull a model, run &lt;code&gt;openclaw onboard&lt;/code&gt; and pick &lt;strong&gt;Ollama&lt;/strong&gt;, then &lt;strong&gt;Local only&lt;/strong&gt;:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight plaintext"&gt;&lt;code&gt;ollama pull gemma4
openclaw onboard
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;The key field is a formality for local hosts: the &lt;a href="https://docs.openclaw.ai/providers/ollama/setup" rel="noopener noreferrer"&gt;Ollama setup page&lt;/a&gt; says loopback URLs "do not need a real bearer token" and "any value works". &lt;code&gt;openclaw gateway install&lt;/code&gt; keeps the gateway running after you log out.&lt;/p&gt;

&lt;p&gt;Two limits. The upstream FAQ answers "Is a local model okay for casual chats?" with "Usually not": small or quantized models truncate context and raise prompt-injection risk, and the managed llama.cpp recipe has an &lt;a href="https://docs.openclaw.ai/gateway/local-models" rel="noopener noreferrer"&gt;8 GiB memory floor&lt;/a&gt;. And the machine is the uptime: sleep, OS updates and network drops are your outages. Right for a desktop with 16 GB or more and an agent that works while you are at the desk. The step-by-step, with every command checked on OpenClaw 2026.9.8, is in &lt;a href="https://aplite.ai/blog/openclaw-hosting-for-free?utm_source=devto&amp;amp;utm_medium=article&amp;amp;utm_campaign=post&amp;amp;utm_content=free-7-ways" rel="noopener noreferrer"&gt;OpenClaw hosting for free&lt;/a&gt;.&lt;/p&gt;

&lt;h2&gt;
  
  
  2. Oracle Cloud Always Free
&lt;/h2&gt;

&lt;p&gt;The most machine anyone gives away: &lt;a href="https://docs.oracle.com/en-us/iaas/Content/FreeTier/freetier.htm" rel="noopener noreferrer"&gt;2 Arm OCPUs and 12 GB of memory&lt;/a&gt;, or two 1 GB micro instances, and 200 GB of block storage, with no end date. Sign-up takes a card and a phone number; "Your credit card will not be charged unless you upgrade your account." You create the VM, open the firewall, run the installer, bring a key or a local model, as in the &lt;a href="https://docs.openclaw.ai/install/oracle" rel="noopener noreferrer"&gt;upstream Oracle guide&lt;/a&gt;.&lt;/p&gt;

&lt;p&gt;Two things end the fun. "Out of host capacity" when creating the Arm shape, sometimes for weeks. And instances that sit idle for 7 days &lt;a href="https://docs.oracle.com/en-us/iaas/Content/FreeTier/freetier_topic-Always_Free_Resources.htm" rel="noopener noreferrer"&gt;may be reclaimed&lt;/a&gt;. Older guides still quote 4 OCPUs and 24 GB; that allowance was halved.&lt;/p&gt;

&lt;h2&gt;
  
  
  3. Google Cloud free tier
&lt;/h2&gt;

&lt;p&gt;One &lt;a href="https://docs.cloud.google.com/free/docs/free-cloud-features" rel="noopener noreferrer"&gt;e2-micro&lt;/a&gt; in us-west1, us-central1 or us-east1 with 30 GB of disk, behind a payment method, no end date. 1 GB of memory is exactly the floor the OpenClaw FAQ names, not its recommendation: capped at 1 GB, the 2026.9.8 gateway settles at about 740 MB resident a minute after start and logs &lt;code&gt;memory pressure: level=warning&lt;/code&gt;. It runs; it does not run a browser tool next to it. The $300 welcome credit lasts 90 days and is a different thing from the always-free micro.&lt;/p&gt;

&lt;h2&gt;
  
  
  4. AWS Free Tier
&lt;/h2&gt;

&lt;p&gt;AWS replaced its always-free micro instance with credits: &lt;a href="https://aws.amazon.com/free/" rel="noopener noreferrer"&gt;$100 on sign-up, up to $200&lt;/a&gt; for completing tasks. The page says the account "closes on its own 6 months after you open it or when your credits run out, whichever comes first". So this is a six-month test, not a home. Your key, your install, any region.&lt;/p&gt;

&lt;h2&gt;
  
  
  5. Railway, with the official template
&lt;/h2&gt;

&lt;p&gt;The OpenClaw docs have a &lt;a href="https://docs.openclaw.ai/install/railway" rel="noopener noreferrer"&gt;Railway install page&lt;/a&gt; and a one-click template that asks only for a gateway token. Railway gives a $5 credit for 30 days with no card; after that the Hobby plan is $5 a month and includes $5 of usage, billed per second beyond it. What you get is a container, not a VM: the volume, the environment variables, updates and &lt;code&gt;openclaw backup create&lt;/code&gt; are yours. Regions: two in the US, Amsterdam, Singapore. Models are your own key. Right for someone who wants a month of testing without touching Linux.&lt;/p&gt;

&lt;h2&gt;
  
  
  6. A managed free plan
&lt;/h2&gt;

&lt;p&gt;Two hosts have a plan with no end date at $0 and no card. Both sleep between uses.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Aplite Free&lt;/strong&gt; (the one I work on) is one agent, the full markdown workspace, approvals for every file change, Telegram from a bot token, budget models from the catalog and 300 welcome credits that do not expire. No API key: models run through one metered account, 1 credit = 1¢, and bringing your own key is not supported. The workspace stops after 6 hours idle and starts again when you open it, usually under a minute, and the Telegram bot answers only while it is awake. No scheduled routines on Free; &lt;a href="https://aplite.ai/docs/plans?utm_source=devto&amp;amp;utm_medium=article&amp;amp;utm_campaign=post&amp;amp;utm_content=free-7-ways" rel="noopener noreferrer"&gt;Pro is $20 a month&lt;/a&gt;. Servers in Germany, EU company.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;&lt;a href="https://www.agent37.com/personal" rel="noopener noreferrer"&gt;Agent37 Free&lt;/a&gt;&lt;/strong&gt; is one agent on 1 vCPU and 4 GB, no card, with a task board and a web terminal. It sleeps after 15 minutes idle, and the model is yours after "$0.20 of the default model to start"; Plus at $9.99 bundles free models.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://www.betterclaw.io/pricing" rel="noopener noreferrer"&gt;BetterClaw&lt;/a&gt; names a Free plan in its FAQ without listing its limits.&lt;/p&gt;

&lt;h2&gt;
  
  
  7. Free hosting, metered tokens
&lt;/h2&gt;

&lt;p&gt;&lt;strong&gt;&lt;a href="https://runware.ai/free-openclaw" rel="noopener noreferrer"&gt;Runware&lt;/a&gt;&lt;/strong&gt; charges nothing for the instance: compute, storage and networking "included at zero charge", "no payment required to deploy", one instance per account during the beta. Every reply draws on a wallet at per-token rates, which a business email starts at $2. An instance unused for 5 days is suspended until you redeploy it, data kept. EU and US residency. Honest and cheap for a bot that talks a few times a day; the wallet is the plan.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;&lt;a href="https://www.ampere.sh/" rel="noopener noreferrer"&gt;Ampere&lt;/a&gt;&lt;/strong&gt; is harder to read. The FAQ says "Free hosting during beta, $10 in AI credits included" and no card; the plan table on the same page lists the entry tier as "Free" at $1 a month with 5,000 credits and a button that reads "Start for $1". Take whichever the sign-up form shows you.&lt;/p&gt;

&lt;h2&gt;
  
  
  Trials are not free plans
&lt;/h2&gt;

&lt;p&gt;Several hosts show up in "free OpenClaw hosting" lists with a trial. A trial is the paid product for a few days: &lt;a href="https://oneclickclaw.io/openclaw" rel="noopener noreferrer"&gt;OneClickClaw&lt;/a&gt; gives 7 days on a real server in Denmark against a Google account, then deletes the server unless you add a payment method (€14.99/mo after); &lt;a href="https://runonflux.com/apps/openclaw/pricing" rel="noopener noreferrer"&gt;Flux&lt;/a&gt; gives new users the first week, then $5.49/mo; &lt;a href="https://rapidclaw.dev/" rel="noopener noreferrer"&gt;RapidClaw&lt;/a&gt; counts 5 messages, then $29/mo; &lt;a href="https://openclawlaunch.com/pricing" rel="noopener noreferrer"&gt;OpenClaw Launch&lt;/a&gt; gives 30 minutes; &lt;a href="https://clawhosters.com/pricing" rel="noopener noreferrer"&gt;ClawHosters&lt;/a&gt; 48 hours with a card on file; ClawTrust 5 days. KiloClaw appears in most lists, but on 19 October every kilo.ai/kiloclaw address redirected to the Kilo Code homepage, so there was nothing to verify.&lt;/p&gt;

&lt;h2&gt;
  
  
  How to pick
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;$0 forever, and you have a 16 GB desktop:&lt;/strong&gt; way 1. Tokens cost nothing and the agent works while the machine does.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;A real server for $0 and you can install:&lt;/strong&gt; way 2. Way 3 if you accept 1 GB and a US region.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;A month of testing, no card, no Linux:&lt;/strong&gt; way 5 for a container you control, way 6 for no install at all.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;A bot that answers a few times a day, no server, no key:&lt;/strong&gt; way 6 or way 7. Both sleep or suspend when idle; that is the price.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Always-on with nothing to maintain:&lt;/strong&gt; none of the seven. The cheapest paid lines, from $4.99 on a 24-month term to $29 with model credits included, are in &lt;a href="https://aplite.ai/blog/managed-openclaw-hosting-compared?utm_source=devto&amp;amp;utm_medium=article&amp;amp;utm_campaign=post&amp;amp;utm_content=free-7-ways" rel="noopener noreferrer"&gt;Managed OpenClaw hosting compared&lt;/a&gt;, 18 hosts in one table.&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;The longer version of this table, 15 hosts and tiers with who takes a card for a plan that never charges it and which trials end by deleting the server, is on &lt;a href="https://aplite.ai/blog/openclaw-free-plan-no-credit-card?utm_source=devto&amp;amp;utm_medium=article&amp;amp;utm_campaign=post&amp;amp;utm_content=free-7-ways" rel="noopener noreferrer"&gt;OpenClaw hosting with a free plan and no credit card&lt;/a&gt;. If the question is the key rather than the server, &lt;a href="https://aplite.ai/blog/openclaw-without-api-key?utm_source=devto&amp;amp;utm_medium=article&amp;amp;utm_campaign=post&amp;amp;utm_content=free-7-ways" rel="noopener noreferrer"&gt;Can you use OpenClaw without an API key?&lt;/a&gt; walks through the subscription logins and the free keys.&lt;/p&gt;

&lt;p&gt;Aplite is an independent service and is not affiliated with the OpenClaw project. Prices on this page move; the date they were read is in the first line.&lt;/p&gt;

</description>
      <category>openclaw</category>
      <category>hosting</category>
      <category>selfhosting</category>
      <category>ai</category>
    </item>
    <item>
      <title>OpenClaw install on Windows and Mac: what the installer does, where it breaks, and the hosted way around it</title>
      <dc:creator>Aplite</dc:creator>
      <pubDate>Wed, 21 Oct 2026 03:38:02 +0000</pubDate>
      <link>https://dev.to/aplite/openclaw-install-on-windows-and-mac-what-the-installer-does-where-it-breaks-and-the-hosted-way-837</link>
      <guid>https://dev.to/aplite/openclaw-install-on-windows-and-mac-what-the-installer-does-where-it-breaks-and-the-hosted-way-837</guid>
      <description>&lt;p&gt;&lt;em&gt;I work on &lt;a href="https://aplite.ai/docs/openclaw?utm_source=devto&amp;amp;utm_medium=article&amp;amp;utm_campaign=post&amp;amp;utm_content=openclaw-install-windows-mac" rel="noopener noreferrer"&gt;Aplite&lt;/a&gt;, a managed OpenClaw host. Everything below about OpenClaw itself comes from the upstream docs, the installer scripts and the issue tracker, and is linked to them; the hosted path at the end is ours, with the trade-off stated.&lt;/em&gt;&lt;/p&gt;

&lt;p&gt;Search for &lt;em&gt;openclaw install windows&lt;/em&gt; or &lt;em&gt;openclaw install mac&lt;/em&gt; and the first page is&lt;br&gt;
vendors: Kimi's guide ends at its own cloud and still says Node 22.14, Parallels wants&lt;br&gt;
you in a VM, hosting blogs want you on their server. The official answer is spread over&lt;br&gt;
five docs pages. This one puts it in order: what the two installer scripts do step by&lt;br&gt;
step, the requirements that changed in September, the errors people hit with their&lt;br&gt;
real text, and a checklist for the end. Everything about OpenClaw is from the&lt;br&gt;
&lt;a href="https://docs.openclaw.ai/install" rel="noopener noreferrer"&gt;OpenClaw docs&lt;/a&gt; and the scripts themselves, read on&lt;br&gt;
20 October 2026 (release 2026.10.5).&lt;/p&gt;
&lt;h2&gt;
  
  
  Requirements, as of October 2026
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Node 24.16 or newer, or 26.1 or newer.&lt;/strong&gt; Node 22, 23 and 25 are unsupported since
2026.9.3, because their &lt;code&gt;node:sqlite&lt;/code&gt; silently truncates text at an embedded NUL
(&lt;a href="https://docs.openclaw.ai/install/node-compatibility" rel="noopener noreferrer"&gt;Node.js compatibility&lt;/a&gt;).
The installer adds Node when it is missing: 26 via Homebrew on a Mac, the LTS line
via winget on Windows.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Windows:&lt;/strong&gt; PowerShell 5 or newer for the script; Windows 10 20H2 or 11 for the
Windows Hub app. WSL2 is optional, not required
(&lt;a href="https://docs.openclaw.ai/platforms/windows" rel="noopener noreferrer"&gt;Windows&lt;/a&gt;).&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Mac:&lt;/strong&gt; the CLI runs where Node runs, macOS 13.5 or newer; the menu bar app needs
macOS 26.2 Tahoe (&lt;a href="https://docs.openclaw.ai/platforms/macos" rel="noopener noreferrer"&gt;macOS app&lt;/a&gt;).&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Git&lt;/strong&gt;, even for an npm install: the installer checks for it so dependencies from
git URLs do not fail with &lt;code&gt;spawn git ENOENT&lt;/code&gt;.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;A model:&lt;/strong&gt; a provider API key, a Claude Code or Codex login, or Ollama. Every
route, including the free ones: &lt;a href="https://aplite.ai/blog/openclaw-without-api-key?utm_source=devto&amp;amp;utm_medium=article&amp;amp;utm_campaign=post&amp;amp;utm_content=openclaw-install-windows-mac" rel="noopener noreferrer"&gt;OpenClaw without an API key&lt;/a&gt;.&lt;/li&gt;
&lt;/ul&gt;
&lt;h2&gt;
  
  
  Which path
&lt;/h2&gt;

&lt;p&gt;On Windows the docs recommend the &lt;strong&gt;Windows Hub&lt;/strong&gt; app&lt;br&gt;
(&lt;a href="https://github.com/openclaw/openclaw-windows-node/releases/latest" rel="noopener noreferrer"&gt;OpenClawCompanion-Setup-x64.exe&lt;/a&gt;,&lt;br&gt;
2026.9.8-1 on 7 October): no administrator rights, and &lt;em&gt;Set up locally&lt;/em&gt; creates its own&lt;br&gt;
WSL distro, &lt;code&gt;OpenClawGateway&lt;/code&gt;. The &lt;strong&gt;PowerShell script&lt;/strong&gt; gives a native CLI and a&lt;br&gt;
gateway kept alive by a Scheduled Task; &lt;strong&gt;WSL2&lt;/strong&gt; runs the Mac/Linux script. On a Mac:&lt;br&gt;
the &lt;strong&gt;menu bar app&lt;/strong&gt; (2026.10.5 ships Apple Silicon, Intel and Universal DMGs) or the&lt;br&gt;
&lt;strong&gt;curl script&lt;/strong&gt;, which installs a LaunchAgent. This page follows the scripts.&lt;/p&gt;
&lt;h2&gt;
  
  
  What install.sh does on a Mac
&lt;/h2&gt;

&lt;p&gt;About five minutes by the &lt;a href="https://docs.openclaw.ai/start/getting-started" rel="noopener noreferrer"&gt;getting started&lt;/a&gt;&lt;br&gt;
page. The flow, from &lt;a href="https://docs.openclaw.ai/install/installer" rel="noopener noreferrer"&gt;installer internals&lt;/a&gt;&lt;br&gt;
and the script:&lt;/p&gt;

&lt;ol&gt;
&lt;li&gt;
&lt;p&gt;&lt;strong&gt;Run it&lt;/strong&gt; in Terminal:&lt;br&gt;
&lt;/p&gt;

&lt;pre class="highlight plaintext"&gt;&lt;code&gt;curl -fsSL https://openclaw.ai/install.sh | bash
&lt;/code&gt;&lt;/pre&gt;


&lt;/li&gt;
&lt;li&gt;&lt;p&gt;&lt;strong&gt;Node.&lt;/strong&gt; The script checks the version and the linked SQLite. If Node is missing it&lt;br&gt;
installs Homebrew first (that needs an Administrator account) and then &lt;code&gt;brew install&lt;br&gt;
node&lt;/code&gt;, which is Node 26. An nvm install is reused when compatible; with nvm present&lt;br&gt;
but no compatible version it offers &lt;code&gt;nvm install 26&lt;/code&gt; and never installs a second nvm.&lt;/p&gt;&lt;/li&gt;
&lt;li&gt;&lt;p&gt;&lt;strong&gt;Git&lt;/strong&gt; via Homebrew if missing, then &lt;code&gt;npm install -g openclaw@latest&lt;/code&gt;, with&lt;br&gt;
&lt;code&gt;--allow-scripts=openclaw&lt;/code&gt; added on npm 11.16 and newer. A failed build pulls in&lt;br&gt;
Xcode Command Line Tools and retries.&lt;/p&gt;&lt;/li&gt;
&lt;li&gt;&lt;p&gt;&lt;strong&gt;Onboarding.&lt;/strong&gt; On a fresh install it hands over to &lt;code&gt;openclaw onboard&lt;/code&gt;; &lt;em&gt;Quick&lt;br&gt;
start&lt;/em&gt; finds a Claude Code or Codex login or a key, checks it with a real&lt;br&gt;
completion and opens the dashboard. On an existing install it runs&lt;br&gt;
&lt;code&gt;openclaw doctor --fix&lt;/code&gt; and restarts the gateway service instead.&lt;/p&gt;&lt;/li&gt;
&lt;li&gt;
&lt;p&gt;&lt;strong&gt;Keep it running.&lt;/strong&gt; Quick start leaves the gateway in the foreground. Press&lt;br&gt;
Ctrl+C, then:&lt;br&gt;
&lt;/p&gt;

&lt;pre class="highlight plaintext"&gt;&lt;code&gt;openclaw gateway install
openclaw gateway status
&lt;/code&gt;&lt;/pre&gt;



&lt;p&gt;The first installs a LaunchAgent; the second should show the gateway listening on&lt;br&gt;
port 18789.&lt;/p&gt;
&lt;/li&gt;
&lt;/ol&gt;
&lt;h2&gt;
  
  
  What install.ps1 does on Windows
&lt;/h2&gt;

&lt;ol&gt;
&lt;li&gt;
&lt;p&gt;&lt;strong&gt;Run it&lt;/strong&gt; in PowerShell (no execution-policy change needed for this form):&lt;br&gt;
&lt;/p&gt;

&lt;pre class="highlight plaintext"&gt;&lt;code&gt;iwr -useb https://openclaw.ai/install.ps1 | iex
&lt;/code&gt;&lt;/pre&gt;


&lt;/li&gt;
&lt;li&gt;&lt;p&gt;&lt;strong&gt;Node.&lt;/strong&gt; If missing: winget, then Chocolatey, then Scoop. If all three are absent&lt;br&gt;
or leave an unsupported version, the script downloads the official Node 26 zip into&lt;br&gt;
&lt;code&gt;%LOCALAPPDATA%\OpenClaw\deps\portable-node&lt;/code&gt; and adds it to your user PATH.&lt;/p&gt;&lt;/li&gt;
&lt;li&gt;&lt;p&gt;&lt;strong&gt;OpenClaw&lt;/strong&gt; by &lt;code&gt;npm install -g&lt;/code&gt;, launched from a temp directory so a shell opened in&lt;br&gt;
&lt;code&gt;C:\&lt;/code&gt; still works. If Git is missing on the git install method, a user-local MinGit&lt;br&gt;
goes under &lt;code&gt;%LOCALAPPDATA%\OpenClaw\deps\portable-git&lt;/code&gt;.&lt;/p&gt;&lt;/li&gt;
&lt;li&gt;&lt;p&gt;&lt;strong&gt;PATH and doctor.&lt;/strong&gt; The bin directory goes on your user PATH; on upgrades it runs&lt;br&gt;
&lt;code&gt;openclaw doctor --fix --non-interactive&lt;/code&gt;. A failure stops with a terminating error&lt;br&gt;
and leaves the window open.&lt;/p&gt;&lt;/li&gt;
&lt;li&gt;
&lt;p&gt;&lt;strong&gt;Onboarding&lt;/strong&gt;, as on the Mac. Then the service:&lt;br&gt;
&lt;/p&gt;

&lt;pre class="highlight plaintext"&gt;&lt;code&gt;openclaw gateway install
openclaw gateway status --json
&lt;/code&gt;&lt;/pre&gt;



&lt;p&gt;This registers a Scheduled Task, "OpenClaw Gateway", that runs as you before login&lt;br&gt;
without storing a password. If task creation is denied you get a Startup-folder&lt;br&gt;
item instead, which only starts after you log in; run the install from an elevated&lt;br&gt;
terminal to get the boot task.&lt;/p&gt;
&lt;/li&gt;
&lt;/ol&gt;
&lt;h2&gt;
  
  
  Where it breaks
&lt;/h2&gt;

&lt;p&gt;The error text as it appears, why, and the fix, with the source.&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;&lt;code&gt;openclaw is not recognized&lt;/code&gt;&lt;/strong&gt; on Windows, &lt;strong&gt;&lt;code&gt;zsh: command not found: openclaw&lt;/code&gt;&lt;/strong&gt; on
a Mac. npm's global bin directory is not on PATH, or the terminal predates the
install. Reopen the terminal; still missing: &lt;code&gt;npm config get prefix&lt;/code&gt; and add that
directory to the user PATH on Windows (no &lt;code&gt;\bin&lt;/code&gt;), on a Mac
&lt;code&gt;export PATH="$(npm prefix -g)/bin:$PATH"&lt;/code&gt; in &lt;code&gt;~/.zshrc&lt;/code&gt;
(&lt;a href="https://docs.openclaw.ai/install/node#troubleshooting" rel="noopener noreferrer"&gt;docs&lt;/a&gt;).&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;&lt;code&gt;npm error spawn git ENOENT&lt;/code&gt;&lt;/strong&gt; on Windows. Git is missing, so a dependency from a
git URL cannot be cloned. Rerun the script so it bootstraps MinGit, or install Git for
Windows and reopen PowerShell
(&lt;a href="https://docs.openclaw.ai/install/installer#troubleshooting" rel="noopener noreferrer"&gt;docs&lt;/a&gt;). January's
&lt;code&gt;node.exe : npm error code ENOENT&lt;/code&gt; was the installer calling npm through &lt;code&gt;node.exe&lt;/code&gt;
(&lt;a href="https://github.com/openclaw/openclaw/issues/3917" rel="noopener noreferrer"&gt;#3917&lt;/a&gt;,
&lt;a href="https://github.com/openclaw/openclaw/issues/11049" rel="noopener noreferrer"&gt;#11049&lt;/a&gt;); gone since.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;&lt;code&gt;Error: Could not find a package manager (winget, choco, or scoop)&lt;/code&gt;&lt;/strong&gt;. Windows
without winget, and the window closed before you could read it
(&lt;a href="https://github.com/openclaw/openclaw/issues/38054" rel="noopener noreferrer"&gt;#38054&lt;/a&gt;, March). Fixed since:
the script downloads portable Node, and the piped form keeps the window open.
If it recurs, install winget from the Store or Node from nodejs.org.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;&lt;code&gt;Node 22.22.0: node:sqlite truncates TEXT at embedded NUL (nodejs/node#61954); use
24.16+/26.1+ or a build with the fix&lt;/code&gt;&lt;/strong&gt;, with a &lt;code&gt;v&lt;/code&gt; before the number on Windows.
Node below the September floor. Let the installer provision Node, or answer &lt;strong&gt;Y&lt;/strong&gt;
to the CLI's &lt;code&gt;Update NodeJS: Y/N [N]:&lt;/code&gt; prompt; a private runtime lands in
&lt;code&gt;~/.openclaw/tools/cli-node&lt;/code&gt;.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;&lt;code&gt;blocked because they are not covered by allowScripts&lt;/code&gt;&lt;/strong&gt;. A manual
&lt;code&gt;npm install -g openclaw&lt;/code&gt; on npm 12. Add &lt;code&gt;--allow-scripts=openclaw&lt;/code&gt;;
&lt;code&gt;npm approve-scripts&lt;/code&gt; does not work for global installs
(&lt;a href="https://docs.openclaw.ai/install#npm-pnpm-or-bun" rel="noopener noreferrer"&gt;docs&lt;/a&gt;).&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;&lt;code&gt;npm.ps1 cannot be loaded because running scripts is disabled on this system&lt;/code&gt;&lt;/strong&gt;,
same for &lt;code&gt;openclaw.ps1&lt;/code&gt;. Running npm or openclaw by hand under the Restricted policy
Windows client editions ship with; npm's &lt;code&gt;.ps1&lt;/code&gt; shims are scripts, the piped
installer is not. &lt;code&gt;Set-ExecutionPolicy RemoteSigned -Scope CurrentUser&lt;/code&gt;, or call
&lt;code&gt;npm.cmd&lt;/code&gt; and &lt;code&gt;openclaw.cmd&lt;/code&gt;
(&lt;a href="https://learn.microsoft.com/en-us/powershell/module/microsoft.powershell.core/about/about_execution_policies" rel="noopener noreferrer"&gt;Microsoft&lt;/a&gt;).&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;&lt;code&gt;You have not agreed to the Xcode license agreements&lt;/code&gt;&lt;/strong&gt;, or the "install the
command line developer tools" dialog on a Mac. npm needs git; without Command Line
Tools, Apple's &lt;code&gt;/usr/bin/git&lt;/code&gt; shim opens the dialog, and the gateway runs git at
every start even with update checks off
(&lt;a href="https://github.com/openclaw/openclaw/issues/31527" rel="noopener noreferrer"&gt;#31527&lt;/a&gt;,
&lt;a href="https://github.com/openclaw/openclaw/issues/158099" rel="noopener noreferrer"&gt;#158099&lt;/a&gt;, open).
&lt;code&gt;xcode-select --install&lt;/code&gt;, finish the dialog, rerun; &lt;code&gt;sudo xcodebuild -license&lt;/code&gt; only
if full Xcode is installed.&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;Two more, after the install rather than during it. On Windows a Scheduled Task that&lt;br&gt;
reports success proves nothing, "A successful task result alone does not prove the&lt;br&gt;
Gateway is healthy": read &lt;code&gt;openclaw gateway status --json&lt;/code&gt; and the gateway log, as the&lt;br&gt;
2026.9.1 update taught (&lt;a href="https://github.com/openclaw/openclaw/issues/137813" rel="noopener noreferrer"&gt;#137813&lt;/a&gt;).&lt;br&gt;
On a Mac an open issue&lt;br&gt;
has the LaunchAgent reach &lt;code&gt;ready&lt;/code&gt;, get SIGTERM a second later and loop, while the same&lt;br&gt;
build is stable in the foreground&lt;br&gt;
(&lt;a href="https://github.com/openclaw/openclaw/issues/144094" rel="noopener noreferrer"&gt;#144094&lt;/a&gt;, 2026.8.2 and 9.3);&lt;br&gt;
&lt;code&gt;openclaw gateway run&lt;/code&gt; tells the two apart. On 2026.9.4 on Windows, read the docs'&lt;br&gt;
&lt;a href="https://docs.openclaw.ai/platforms/windows#updating-from-2026-9-4" rel="noopener noreferrer"&gt;note on updating&lt;/a&gt;&lt;br&gt;
before &lt;code&gt;openclaw update&lt;/code&gt;.&lt;/p&gt;

&lt;h2&gt;
  
  
  After the install: a checklist
&lt;/h2&gt;

&lt;ol&gt;
&lt;li&gt;
&lt;code&gt;openclaw --version&lt;/code&gt; prints 2026.10.5 or whatever you asked for.&lt;/li&gt;
&lt;li&gt;
&lt;code&gt;openclaw doctor&lt;/code&gt; has nothing red; &lt;code&gt;openclaw doctor --fix&lt;/code&gt; repairs what it can.&lt;/li&gt;
&lt;li&gt;
&lt;code&gt;openclaw gateway status&lt;/code&gt; shows the gateway on port 18789, from a new terminal.&lt;/li&gt;
&lt;li&gt;
&lt;code&gt;openclaw dashboard&lt;/code&gt; opens the Control UI and a typed message gets a reply.&lt;/li&gt;
&lt;li&gt;The gateway is bound to &lt;code&gt;127.0.0.1&lt;/code&gt;, which is the default: do not forward 18789
on your router (&lt;a href="https://docs.openclaw.ai/gateway/security" rel="noopener noreferrer"&gt;security&lt;/a&gt;).&lt;/li&gt;
&lt;li&gt;A first channel: &lt;a href="https://aplite.ai/blog/openclaw-telegram-setup?utm_source=devto&amp;amp;utm_medium=article&amp;amp;utm_campaign=post&amp;amp;utm_content=openclaw-install-windows-mac" rel="noopener noreferrer"&gt;OpenClaw in Telegram in five minutes&lt;/a&gt;.
An unknown sender gets a pairing code until you approve them.&lt;/li&gt;
&lt;li&gt;The machine stays on. Sleep, a closed lid and OS updates stop the gateway; the
upstream FAQ says to use a VPS after disconnects
(&lt;a href="https://docs.openclaw.ai/help/faq-first-run/providers-and-hosting" rel="noopener noreferrer"&gt;providers and hosting&lt;/a&gt;).&lt;/li&gt;
&lt;li&gt;When something is off, &lt;code&gt;openclaw triage&lt;/code&gt; runs read-only checks and writes a
diagnosis for Claude Code, Codex or the built-in agent.&lt;/li&gt;
&lt;/ol&gt;

&lt;h2&gt;
  
  
  Without installing anything
&lt;/h2&gt;

&lt;p&gt;&lt;a href="https://aplite.ai/docs/openclaw?utm_source=devto&amp;amp;utm_medium=article&amp;amp;utm_campaign=post&amp;amp;utm_content=openclaw-install-windows-mac" rel="noopener noreferrer"&gt;Aplite&lt;/a&gt; runs a standard OpenClaw instance for you: one gateway in its own&lt;br&gt;
container on our server in Nuremberg, currently 2026.9.8, set up from the browser.&lt;/p&gt;

&lt;ol&gt;
&lt;li&gt;Create an account with email or Google; no card on the Free plan.&lt;/li&gt;
&lt;li&gt;Pick an agent and press &lt;strong&gt;Deploy&lt;/strong&gt;; the workspace opens about a minute later.&lt;/li&gt;
&lt;li&gt;Ask. There is no Node, no PATH, no key to paste: models come through one metered
account, billed in credits, 1 credit = 1¢, token usage under each reply.&lt;/li&gt;
&lt;/ol&gt;

&lt;p&gt;What a managed instance gives up, the shell, raw &lt;code&gt;openclaw.json&lt;/code&gt;, local models and&lt;br&gt;
day-one releases, is in &lt;a href="https://aplite.ai/blog/managed-openclaw-vs-self-hosting?utm_source=devto&amp;amp;utm_medium=article&amp;amp;utm_campaign=post&amp;amp;utm_content=openclaw-install-windows-mac" rel="noopener noreferrer"&gt;Managed OpenClaw vs self-hosting&lt;/a&gt;;&lt;br&gt;
other hosts are in &lt;a href="https://aplite.ai/blog/managed-openclaw-hosting-compared?utm_source=devto&amp;amp;utm_medium=article&amp;amp;utm_campaign=post&amp;amp;utm_content=openclaw-install-windows-mac" rel="noopener noreferrer"&gt;Managed OpenClaw hosting compared&lt;/a&gt;.&lt;br&gt;
Aplite is independent and not affiliated with the OpenClaw project.&lt;/p&gt;

&lt;h2&gt;
  
  
  FAQ
&lt;/h2&gt;

&lt;h3&gt;
  
  
  Does OpenClaw run on Windows?
&lt;/h3&gt;

&lt;p&gt;Yes, three ways: the Windows Hub desktop app, which creates its own WSL distro; the&lt;br&gt;
PowerShell installer, a native CLI and a gateway kept alive by a Scheduled Task; or&lt;br&gt;
WSL2 with the Linux script. The docs recommend the Hub for desktop use.&lt;/p&gt;

&lt;h3&gt;
  
  
  Do I need WSL2 to install OpenClaw on Windows?
&lt;/h3&gt;

&lt;p&gt;No. The PowerShell script installs a native gateway. WSL2 is the most Linux-compatible&lt;br&gt;
option and what the Hub uses underneath, but it is not required.&lt;/p&gt;

&lt;h3&gt;
  
  
  Which Node version does OpenClaw need?
&lt;/h3&gt;

&lt;p&gt;Node 24.16 or newer, or 26.1 or newer; Node 26 is recommended. Node 22, 23 and 25 are&lt;br&gt;
unsupported since release 2026.9.3. The installer provisions a supported Node when&lt;br&gt;
yours is missing or too old.&lt;/p&gt;

&lt;h3&gt;
  
  
  Does OpenClaw work on an Intel Mac or an older macOS?
&lt;/h3&gt;

&lt;p&gt;The CLI and gateway run wherever Node 24 or 26 runs, which Node builds for macOS 13.5&lt;br&gt;
and newer, on Intel and Apple Silicon. The menu bar app needs macOS 26.2 Tahoe; on an&lt;br&gt;
older Mac use the script.&lt;/p&gt;

&lt;h3&gt;
  
  
  Why does PowerShell say openclaw is not recognized after the install?
&lt;/h3&gt;

&lt;p&gt;The global npm bin directory is not on the PATH of that window. Open a new PowerShell&lt;br&gt;
first; if it still fails, run &lt;code&gt;npm config get prefix&lt;/code&gt; and add that directory to your&lt;br&gt;
user PATH, then reopen.&lt;/p&gt;

&lt;h3&gt;
  
  
  Can I try OpenClaw without installing it?
&lt;/h3&gt;

&lt;p&gt;Yes. A managed host runs the gateway for you; on Aplite that is an account, Deploy and&lt;br&gt;
a chat, on the Free plan without a card. The trade-offs are above.&lt;/p&gt;

</description>
      <category>openclaw</category>
      <category>windows</category>
      <category>macos</category>
      <category>tutorial</category>
    </item>
    <item>
      <title>OpenClaw in Telegram in five minutes: setup, pairing, hosted or self-hosted</title>
      <dc:creator>Aplite</dc:creator>
      <pubDate>Wed, 21 Oct 2026 03:37:45 +0000</pubDate>
      <link>https://dev.to/aplite/openclaw-in-telegram-in-five-minutes-setup-pairing-hosted-or-self-hosted-2dcl</link>
      <guid>https://dev.to/aplite/openclaw-in-telegram-in-five-minutes-setup-pairing-hosted-or-self-hosted-2dcl</guid>
      <description>&lt;p&gt;&lt;em&gt;I work on &lt;a href="https://aplite.ai/docs/openclaw?utm_source=devto&amp;amp;utm_medium=article&amp;amp;utm_campaign=post&amp;amp;utm_content=openclaw-telegram-setup" rel="noopener noreferrer"&gt;Aplite&lt;/a&gt;, a managed OpenClaw host. The self-hosted path below is the upstream one, run on OpenClaw 2026.9.8 and linked to the docs at every step; the hosted path is ours, with the trade-off stated.&lt;/em&gt;&lt;/p&gt;

&lt;p&gt;Search for &lt;em&gt;openclaw telegram setup&lt;/em&gt; and you get a dozen walkthroughs by hosting companies,&lt;br&gt;
each ending at their own control panel, and the upstream documentation, which is correct but&lt;br&gt;
spread over four pages: setup, pairing, access control, troubleshooting. The actual job is&lt;br&gt;
short. A Telegram bot is a token from &lt;a class="mentioned-user" href="https://dev.to/botfather"&gt;@botfather&lt;/a&gt;; OpenClaw needs that token and then an&lt;br&gt;
approval for every chat that may talk to it. That is the whole thing, and it takes about&lt;br&gt;
five minutes whether the gateway runs on your machine or on someone else's.&lt;/p&gt;

&lt;p&gt;Below, both paths on OpenClaw 2026.9.8: the self-hosted one with every command, checked&lt;br&gt;
against the &lt;a href="https://docs.openclaw.ai/channels/telegram/setup" rel="noopener noreferrer"&gt;upstream docs&lt;/a&gt; and run in a&lt;br&gt;
fresh container of that version, then the hosted one,&lt;br&gt;
where the commands become two buttons.&lt;/p&gt;
&lt;h2&gt;
  
  
  What you need
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;A bot token.&lt;/strong&gt; Free, from &lt;a href="https://t.me/BotFather" rel="noopener noreferrer"&gt;@BotFather&lt;/a&gt; inside Telegram. Check
the handle is exactly &lt;code&gt;@BotFather&lt;/code&gt;; there are impostors.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;A gateway that stays up.&lt;/strong&gt; The Telegram channel long-polls Telegram's API from the
gateway, so a bot only answers while the gateway runs. A laptop that sleeps stops it.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;A model.&lt;/strong&gt; The bot token is not a model key. Self-hosted, you still need a provider or
a local model (&lt;a href="https://aplite.ai/blog/openclaw-without-api-key?utm_source=devto&amp;amp;utm_medium=article&amp;amp;utm_campaign=post&amp;amp;utm_content=openclaw-telegram-setup" rel="noopener noreferrer"&gt;the ways without an API key&lt;/a&gt;); hosted,
the models are included.&lt;/li&gt;
&lt;/ul&gt;
&lt;h2&gt;
  
  
  Self-hosted: six steps
&lt;/h2&gt;

&lt;ol&gt;
&lt;li&gt;&lt;p&gt;&lt;strong&gt;Create the bot.&lt;/strong&gt; In Telegram, open &lt;a class="mentioned-user" href="https://dev.to/botfather"&gt;@botfather&lt;/a&gt;, send &lt;code&gt;/newbot&lt;/code&gt;, answer with a display&lt;br&gt;
name and a username ending in &lt;code&gt;bot&lt;/code&gt;. BotFather replies with the token, shaped like&lt;br&gt;
&lt;code&gt;123456789:AAF…&lt;/code&gt;. Keep it private: whoever has it can run your bot until you regenerate&lt;br&gt;
it with &lt;code&gt;/revoke&lt;/code&gt;.&lt;/p&gt;&lt;/li&gt;
&lt;li&gt;
&lt;p&gt;&lt;strong&gt;Install OpenClaw&lt;/strong&gt; if it is not there yet. On macOS, Linux or WSL2:&lt;br&gt;
&lt;/p&gt;

&lt;pre class="highlight plaintext"&gt;&lt;code&gt;curl -fsSL https://openclaw.ai/install.sh | bash
openclaw onboard
&lt;/code&gt;&lt;/pre&gt;



&lt;p&gt;Onboarding picks the model provider and starts the gateway; the longer version is in&lt;br&gt;
&lt;a href="https://aplite.ai/blog/openclaw-hosting-for-free?utm_source=devto&amp;amp;utm_medium=article&amp;amp;utm_campaign=post&amp;amp;utm_content=openclaw-telegram-setup#the-0-path-step-by-step" rel="noopener noreferrer"&gt;OpenClaw hosting for free&lt;/a&gt;.&lt;/p&gt;
&lt;/li&gt;
&lt;li&gt;
&lt;p&gt;&lt;strong&gt;Give OpenClaw the token.&lt;/strong&gt; One command:&lt;br&gt;
&lt;/p&gt;

&lt;pre class="highlight plaintext"&gt;&lt;code&gt;openclaw channels add --channel telegram --token 123456789:AAF…
&lt;/code&gt;&lt;/pre&gt;



&lt;p&gt;It prints &lt;code&gt;Added Telegram account "default".&lt;/code&gt; and writes the channel into&lt;br&gt;
&lt;code&gt;~/.openclaw/openclaw.json&lt;/code&gt;:&lt;br&gt;
&lt;/p&gt;

&lt;pre class="highlight plaintext"&gt;&lt;code&gt;"channels": {
  "telegram": {
    "enabled": true,
    "botToken": "123456789:AAF…"
  }
}
&lt;/code&gt;&lt;/pre&gt;



&lt;p&gt;The same block typed by hand works too, as does the &lt;code&gt;TELEGRAM_BOT_TOKEN&lt;/code&gt; environment&lt;br&gt;
variable for the default account; a &lt;code&gt;tokenFile&lt;/code&gt; beats &lt;code&gt;botToken&lt;/code&gt;, which beats the env.&lt;br&gt;
There is no &lt;code&gt;openclaw channels login telegram&lt;/code&gt; step: unlike WhatsApp with its QR&lt;br&gt;
login, Telegram needs nothing but the token&lt;br&gt;
(&lt;a href="https://docs.openclaw.ai/channels/telegram/setup" rel="noopener noreferrer"&gt;setup docs&lt;/a&gt;).&lt;/p&gt;
&lt;/li&gt;
&lt;li&gt;
&lt;p&gt;&lt;strong&gt;Make sure the gateway is running.&lt;/strong&gt; A running gateway hot-reloads the new channel. If&lt;br&gt;
it is not running, start it in the foreground with &lt;code&gt;openclaw gateway&lt;/code&gt;, or as the service&lt;br&gt;
that &lt;code&gt;openclaw gateway install&lt;/code&gt; set up (&lt;a href="https://docs.openclaw.ai/gateway" rel="noopener noreferrer"&gt;gateway docs&lt;/a&gt;):&lt;br&gt;
&lt;/p&gt;

&lt;pre class="highlight plaintext"&gt;&lt;code&gt;openclaw gateway start
&lt;/code&gt;&lt;/pre&gt;



&lt;p&gt;Then check the channel:&lt;br&gt;
&lt;/p&gt;

&lt;pre class="highlight plaintext"&gt;&lt;code&gt;openclaw channels status --probe
&lt;/code&gt;&lt;/pre&gt;



&lt;p&gt;The Telegram line reads &lt;code&gt;Telegram default: enabled, configured, mode:polling,&lt;br&gt;
token:***&lt;/code&gt;; with &lt;code&gt;--probe&lt;/code&gt; the gateway also calls Telegram with the token. A wrong&lt;br&gt;
token fails at gateway startup as &lt;code&gt;getMe returned 401&lt;/code&gt;.&lt;/p&gt;
&lt;/li&gt;
&lt;li&gt;&lt;p&gt;&lt;strong&gt;Send &lt;code&gt;/start&lt;/code&gt; to your bot.&lt;/strong&gt; A Telegram bot cannot write to you first. On the first&lt;br&gt;
message from an unknown chat, OpenClaw's default DM policy, &lt;code&gt;pairing&lt;/code&gt;, does not answer&lt;br&gt;
the question; it replies with a pairing code instead: eight characters, uppercase, no&lt;br&gt;
&lt;code&gt;0&lt;/code&gt;, &lt;code&gt;O&lt;/code&gt;, &lt;code&gt;1&lt;/code&gt; or &lt;code&gt;I&lt;/code&gt;, valid for one hour, at most three pending per bot&lt;br&gt;
(&lt;a href="https://docs.openclaw.ai/channels/pairing" rel="noopener noreferrer"&gt;pairing docs&lt;/a&gt;).&lt;/p&gt;&lt;/li&gt;
&lt;li&gt;
&lt;p&gt;&lt;strong&gt;Approve the chat.&lt;/strong&gt; On the server:&lt;br&gt;
&lt;/p&gt;

&lt;pre class="highlight plaintext"&gt;&lt;code&gt;openclaw pairing list telegram
openclaw pairing approve telegram K7QX2MPB
&lt;/code&gt;&lt;/pre&gt;



&lt;p&gt;Add &lt;code&gt;--notify&lt;/code&gt; and the bot tells the requester they are in. From now on that chat talks&lt;br&gt;
to the agent; the approval lives in &lt;code&gt;~/.openclaw/state/openclaw.sqlite&lt;/code&gt;, and the first&lt;br&gt;
approved sender becomes the command owner if none was set&lt;br&gt;
(&lt;a href="https://docs.openclaw.ai/cli/pairing" rel="noopener noreferrer"&gt;CLI reference&lt;/a&gt;). Anyone else who finds the bot&lt;br&gt;
gets a code and silence until you approve them, which is the point.&lt;/p&gt;
&lt;/li&gt;
&lt;/ol&gt;

&lt;p&gt;Two things people do next, both optional. &lt;strong&gt;Lock it to yourself&lt;/strong&gt;: set &lt;code&gt;dmPolicy&lt;/code&gt; to&lt;br&gt;
&lt;code&gt;allowlist&lt;/code&gt; and put your numeric Telegram user ID in &lt;code&gt;allowFrom&lt;/code&gt;. The ID is in the bot's&lt;br&gt;
pairing reply, in &lt;code&gt;openclaw logs --follow&lt;/code&gt; as &lt;code&gt;senderUserId&lt;/code&gt;, or at&lt;br&gt;
&lt;code&gt;https://api.telegram.org/bot&amp;lt;token&amp;gt;/getUpdates&lt;/code&gt;; &lt;code&gt;@username&lt;/code&gt; entries are not accepted&lt;br&gt;
(&lt;a href="https://docs.openclaw.ai/channels/telegram/access-control" rel="noopener noreferrer"&gt;access control&lt;/a&gt;). &lt;strong&gt;Add it to a&lt;br&gt;
group&lt;/strong&gt;: groups default to &lt;code&gt;groupPolicy: "allowlist"&lt;/code&gt;, so a group is silent until its chat ID&lt;br&gt;
(a &lt;code&gt;-100…&lt;/code&gt; number for supergroups) is listed under &lt;code&gt;channels.telegram.groups&lt;/code&gt;; the bot then&lt;br&gt;
answers only when mentioned unless &lt;code&gt;requireMention&lt;/code&gt; is off, and reading unmentioned messages&lt;br&gt;
needs Privacy Mode disabled in BotFather (&lt;code&gt;/setprivacy&lt;/code&gt;) or admin rights, then a remove and&lt;br&gt;
re-add. &lt;code&gt;/whoami@your_bot&lt;/code&gt; in the group confirms both IDs.&lt;/p&gt;

&lt;h2&gt;
  
  
  Hosted: the same bot in three steps
&lt;/h2&gt;

&lt;p&gt;&lt;a href="https://aplite.ai/docs/openclaw?utm_source=devto&amp;amp;utm_medium=article&amp;amp;utm_campaign=post&amp;amp;utm_content=openclaw-telegram-setup" rel="noopener noreferrer"&gt;Aplite&lt;/a&gt; runs a standard OpenClaw instance for you, so the gateway is already&lt;br&gt;
up and the config is a form. Same bot, same pairing rule, no server:&lt;/p&gt;

&lt;ol&gt;
&lt;li&gt;&lt;p&gt;&lt;strong&gt;Create a workspace.&lt;/strong&gt; Email or Google, no card on the Free plan. Pick an agent, press&lt;br&gt;
Deploy, and about a minute later the workspace opens with its own OpenClaw 2026.9.8 in&lt;br&gt;
its own container.&lt;/p&gt;&lt;/li&gt;
&lt;li&gt;&lt;p&gt;&lt;strong&gt;Connect the bot.&lt;/strong&gt; Open &lt;strong&gt;Skills → Channels → Telegram&lt;/strong&gt; and press &lt;strong&gt;Connect&lt;/strong&gt;. Paste&lt;br&gt;
the token from &lt;a class="mentioned-user" href="https://dev.to/botfather"&gt;@botfather&lt;/a&gt;, press &lt;strong&gt;Validate&lt;/strong&gt;: Aplite calls Telegram's &lt;code&gt;getMe&lt;/code&gt; and shows&lt;br&gt;
&lt;code&gt;Connected to @your_bot&lt;/code&gt;. Press &lt;strong&gt;Connect&lt;/strong&gt; again; the workspace restarts to pick up the&lt;br&gt;
bot, about a minute. Under &lt;em&gt;Show advanced&lt;/em&gt; sit the same knobs the config file has: DM&lt;br&gt;
policy (pairing by default), group policy (allowlist), chat IDs, require @-mention.&lt;/p&gt;&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/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fi899zeqm9ilc4fiao9zd.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%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fi899zeqm9ilc4fiao9zd.webp" alt="Skills → Channels → Telegram: the card is one Connect button away from a bot." width="800" height="500"&gt;&lt;/a&gt;&lt;br&gt;
&lt;em&gt;Skills → Channels → Telegram: the card is one Connect button away from a bot.&lt;/em&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%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fowyoob0oaxs4qy3ad5pr.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%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fowyoob0oaxs4qy3ad5pr.webp" alt="Paste the token, Validate, Connect: the dialog does what channels add does." width="800" height="500"&gt;&lt;/a&gt;&lt;br&gt;
&lt;em&gt;Paste the token, Validate, Connect: the dialog does what channels add does.&lt;/em&gt;&lt;/p&gt;

&lt;ol&gt;
&lt;li&gt;
&lt;strong&gt;Say &lt;code&gt;/start&lt;/code&gt;, then Approve.&lt;/strong&gt; Send &lt;code&gt;/start&lt;/code&gt; to the bot from your phone. A pairing
request appears on the same Telegram card within seconds, with the sender's name and the
code; press &lt;strong&gt;Approve&lt;/strong&gt;. Codes expire after one hour, so if you were slow, send &lt;code&gt;/start&lt;/code&gt;
again. Only chats you approve can talk to your agent.&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/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fillu9sz4re4t5azmcgh0.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%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fillu9sz4re4t5azmcgh0.webp" alt="The advanced block: pairing, allowlist, open or disabled for DMs; allowlist for groups; the mention rule." width="504" height="1084"&gt;&lt;/a&gt;&lt;br&gt;
&lt;em&gt;The advanced block: pairing, allowlist, open or disabled for DMs; allowlist for groups; the mention rule.&lt;/em&gt;&lt;/p&gt;

&lt;p&gt;That is the setup. The bot answers in Telegram with the same agent that answers in the&lt;br&gt;
browser, routines can &lt;a href="https://aplite.ai/blog/ai-assistant-in-telegram?utm_source=devto&amp;amp;utm_medium=article&amp;amp;utm_campaign=post&amp;amp;utm_content=openclaw-telegram-setup" rel="noopener noreferrer"&gt;deliver to it&lt;/a&gt;, and the toggle on the&lt;br&gt;
Telegram card pauses the channel without dropping the token. One honest line about the Free&lt;br&gt;
plan: the workspace sleeps after 6 hours idle and wakes on your next visit, and while it&lt;br&gt;
sleeps the bot does not answer, so a bot that answers at any hour needs&lt;br&gt;
&lt;a href="https://aplite.ai/docs/plans?utm_source=devto&amp;amp;utm_medium=article&amp;amp;utm_campaign=post&amp;amp;utm_content=openclaw-telegram-setup" rel="noopener noreferrer"&gt;Pro or Power&lt;/a&gt;. Aplite is an independent service and is not affiliated with the&lt;br&gt;
OpenClaw project.&lt;/p&gt;

&lt;h2&gt;
  
  
  If the bot stays silent
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Nothing at all after &lt;code&gt;/start&lt;/code&gt;.&lt;/strong&gt; It is waiting for approval. Run
&lt;code&gt;openclaw pairing list telegram&lt;/code&gt;, approve the code, or approve on the Telegram card. A
code older than an hour is gone; send &lt;code&gt;/start&lt;/code&gt; again
(&lt;a href="https://docs.openclaw.ai/channels/troubleshooting" rel="noopener noreferrer"&gt;channel troubleshooting&lt;/a&gt;).&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;&lt;code&gt;getMe returned 401&lt;/code&gt; or &lt;code&gt;deleteWebhook 401 Unauthorized&lt;/code&gt; at startup.&lt;/strong&gt; The token is
wrong or revoked. Copy it again from BotFather and update &lt;code&gt;botToken&lt;/code&gt;, &lt;code&gt;tokenFile&lt;/code&gt; or
&lt;code&gt;TELEGRAM_BOT_TOKEN&lt;/code&gt;
(&lt;a href="https://docs.openclaw.ai/channels/telegram/troubleshooting" rel="noopener noreferrer"&gt;Telegram troubleshooting&lt;/a&gt;).&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Answers in DMs, silent in a group.&lt;/strong&gt; The group is not in &lt;code&gt;channels.telegram.groups&lt;/code&gt;, or
Privacy Mode is on while you expect unmentioned messages: &lt;code&gt;/setprivacy&lt;/code&gt; → Disable, then
remove and re-add the bot. Hosted, the same two settings are in the advanced block.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Worked yesterday, silent today.&lt;/strong&gt; Self-hosted: &lt;code&gt;openclaw gateway status&lt;/code&gt; should say
&lt;code&gt;Runtime: running&lt;/code&gt;; a laptop that slept or a VPS that rebooted without
&lt;code&gt;openclaw gateway install&lt;/code&gt; stops polling. Hosted on Free: the workspace is asleep; open it
once.&lt;/li&gt;
&lt;/ul&gt;

&lt;h2&gt;
  
  
  FAQ
&lt;/h2&gt;

&lt;h3&gt;
  
  
  Does OpenClaw need a server to work in Telegram?
&lt;/h3&gt;

&lt;p&gt;Yes. The gateway polls Telegram for new messages, so something has to keep running: your&lt;br&gt;
machine while it is awake, a VPS, or a managed instance. A message sent while the gateway&lt;br&gt;
is down is not answered until it is back.&lt;/p&gt;

&lt;h3&gt;
  
  
  Can more than one person use the bot?
&lt;/h3&gt;

&lt;p&gt;Yes. Each person sends &lt;code&gt;/start&lt;/code&gt;, gets their own pairing code, and you approve each one; an&lt;br&gt;
unapproved sender never reaches the agent. For a shared group, list its chat ID under&lt;br&gt;
&lt;code&gt;groups&lt;/code&gt; or switch the group policy to open.&lt;/p&gt;

&lt;h3&gt;
  
  
  Do I need an API key for the Telegram bot?
&lt;/h3&gt;

&lt;p&gt;Not for Telegram: the bot token from &lt;a class="mentioned-user" href="https://dev.to/botfather"&gt;@botfather&lt;/a&gt; is free. The model behind the bot is the&lt;br&gt;
usual question. Self-hosted, that is a provider key, a subscription login or a local model&lt;br&gt;
(&lt;a href="https://aplite.ai/blog/openclaw-without-api-key?utm_source=devto&amp;amp;utm_medium=article&amp;amp;utm_campaign=post&amp;amp;utm_content=openclaw-telegram-setup" rel="noopener noreferrer"&gt;every option&lt;/a&gt;); on Aplite the models are metered and&lt;br&gt;
included.&lt;/p&gt;

&lt;h3&gt;
  
  
  Why does the bot answer my question with a code?
&lt;/h3&gt;

&lt;p&gt;That is pairing, OpenClaw's default DM policy: the first message from an unknown chat gets&lt;br&gt;
a code, not an answer, until the operator approves it with&lt;br&gt;
&lt;code&gt;openclaw pairing approve telegram &amp;lt;CODE&amp;gt;&lt;/code&gt; or the Approve button. Set &lt;code&gt;dmPolicy&lt;/code&gt; to &lt;code&gt;open&lt;/code&gt;&lt;br&gt;
only if you want anyone who finds the bot to command your agent.&lt;/p&gt;

</description>
      <category>openclaw</category>
      <category>telegram</category>
      <category>tutorial</category>
      <category>ai</category>
    </item>
    <item>
      <title>What is OpenClaw, in simple terms (and how to try it without installing anything)</title>
      <dc:creator>Aplite</dc:creator>
      <pubDate>Wed, 21 Oct 2026 03:37:26 +0000</pubDate>
      <link>https://dev.to/aplite/what-is-openclaw-in-simple-terms-and-how-to-try-it-without-installing-anything-3k76</link>
      <guid>https://dev.to/aplite/what-is-openclaw-in-simple-terms-and-how-to-try-it-without-installing-anything-3k76</guid>
      <description>&lt;p&gt;&lt;em&gt;I work on &lt;a href="https://aplite.ai/docs/openclaw?utm_source=devto&amp;amp;utm_medium=article&amp;amp;utm_campaign=post&amp;amp;utm_content=what-is-openclaw" rel="noopener noreferrer"&gt;Aplite&lt;/a&gt;, a managed OpenClaw host. Everything below about OpenClaw itself comes from the upstream docs and is linked to them; the hosted path at the end is ours, with the trade-off stated.&lt;/em&gt;&lt;/p&gt;

&lt;p&gt;Search for &lt;em&gt;what is openclaw&lt;/em&gt; and the first page answers three different questions.&lt;br&gt;
Wikipedia tells the story: an Austrian developer, four names in two months, a star count.&lt;br&gt;
The official docs explain the architecture: trust boundaries, runtimes, a WebSocket&lt;br&gt;
protocol. Hosting companies explain it on a page that ends at their control panel. None&lt;br&gt;
of them says in plain words what the thing is, what the five words in every tutorial&lt;br&gt;
mean, and what it will and will not do. This page does, with the terms as the&lt;br&gt;
&lt;a href="https://docs.openclaw.ai" rel="noopener noreferrer"&gt;OpenClaw docs&lt;/a&gt; use them, read on 20 October 2026.&lt;/p&gt;
&lt;h2&gt;
  
  
  The short version
&lt;/h2&gt;

&lt;p&gt;OpenClaw is an open-source program that you run on a computer you control, and that turns&lt;br&gt;
an AI model into an assistant you message from the chat apps you already use. The docs'&lt;br&gt;
own line: "an open-source AI assistant that runs on your own hardware and meets you in&lt;br&gt;
every chat app you already use."&lt;/p&gt;

&lt;p&gt;&lt;em&gt;Open source&lt;/em&gt;: the code is on &lt;a href="https://github.com/openclaw/openclaw" rel="noopener noreferrer"&gt;GitHub&lt;/a&gt; under the&lt;br&gt;
MIT license, about 393,000 stars on 20 October 2026, stewarded by the OpenClaw&lt;br&gt;
Foundation, a US non-profit that sells nothing ("No subscription. No hosted tier. No&lt;br&gt;
token." on &lt;a href="https://openclaw.ai" rel="noopener noreferrer"&gt;openclaw.ai&lt;/a&gt;). &lt;em&gt;You run it&lt;/em&gt;: on a laptop, a Mac mini or&lt;br&gt;
a rented server, and only while that machine is on. &lt;em&gt;An AI model&lt;/em&gt;: OpenClaw contains&lt;br&gt;
none. It connects to one, from Anthropic, OpenAI, Google and others through a key or a&lt;br&gt;
subscription login, or a local model through Ollama. The model thinks; OpenClaw is&lt;br&gt;
everything around it: the chat connections, the memory on disk, the tools, the schedule.&lt;br&gt;
The &lt;a href="https://github.com/openclaw/openclaw/releases" rel="noopener noreferrer"&gt;current release&lt;/a&gt; is 2026.10.5.&lt;/p&gt;
&lt;h2&gt;
  
  
  The five words
&lt;/h2&gt;

&lt;p&gt;&lt;strong&gt;Gateway.&lt;/strong&gt; The one process that runs all the time: it holds the connections to your&lt;br&gt;
chat apps, serves the browser dashboard, runs the scheduler and talks to the model, on&lt;br&gt;
port 18789 of the machine itself by default&lt;br&gt;
(&lt;a href="https://docs.openclaw.ai/concepts/architecture" rel="noopener noreferrer"&gt;architecture&lt;/a&gt;). When it is down,&lt;br&gt;
nothing answers; a message sent while your laptop sleeps waits until it wakes. That is&lt;br&gt;
why "where does the gateway run" is the first practical question.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Agent.&lt;/strong&gt; A configured assistant with its own folder, default model and history. The&lt;br&gt;
folder, the workspace, holds plain markdown you can edit: &lt;code&gt;AGENTS.md&lt;/code&gt; for operating&lt;br&gt;
instructions, &lt;code&gt;SOUL.md&lt;/code&gt; for persona, &lt;code&gt;USER.md&lt;/code&gt; for who you are, &lt;code&gt;MEMORY.md&lt;/code&gt; for long-term&lt;br&gt;
memory (&lt;a href="https://docs.openclaw.ai/concepts/agent" rel="noopener noreferrer"&gt;agent&lt;/a&gt;). A fresh install has one agent,&lt;br&gt;
&lt;code&gt;main&lt;/code&gt;; more can be added, with different chats routed to each&lt;br&gt;
(&lt;a href="https://docs.openclaw.ai/concepts/multi-agent" rel="noopener noreferrer"&gt;multi-agent&lt;/a&gt;).&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Session.&lt;/strong&gt; A conversation. By default every direct message from every channel lands in&lt;br&gt;
one rolling conversation, the main session, so what you said in Telegram is known when&lt;br&gt;
you continue in the browser; groups get their own, and scheduled jobs start fresh ones&lt;br&gt;
(&lt;a href="https://docs.openclaw.ai/concepts/session" rel="noopener noreferrer"&gt;sessions&lt;/a&gt;).&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Skills.&lt;/strong&gt; Instructions for a specific job: a folder with a &lt;code&gt;SKILL.md&lt;/code&gt; that names the&lt;br&gt;
skill and says when and how to use which tools&lt;br&gt;
(&lt;a href="https://docs.openclaw.ai/tools/skills" rel="noopener noreferrer"&gt;skills&lt;/a&gt;). Some ship with OpenClaw, some you&lt;br&gt;
write, some you install from &lt;a href="https://clawhub.ai" rel="noopener noreferrer"&gt;ClawHub&lt;/a&gt;, the public registry, with&lt;br&gt;
&lt;code&gt;openclaw skills install @owner/&amp;lt;slug&amp;gt;&lt;/code&gt;; the docs call a third-party skill untrusted&lt;br&gt;
code, to be read first.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Channels.&lt;/strong&gt; Where messages come in and go out. Telegram and the browser's WebChat are in&lt;br&gt;
the core install; Discord, Slack, WhatsApp, Signal, iMessage, Matrix, Microsoft Teams,&lt;br&gt;
Google Chat and a dozen more are official plugins&lt;br&gt;
(&lt;a href="https://docs.openclaw.ai/channels" rel="noopener noreferrer"&gt;channels&lt;/a&gt;). An unknown sender gets a pairing code&lt;br&gt;
instead of an answer until you approve them.&lt;/p&gt;

&lt;p&gt;Four more, briefly: &lt;strong&gt;tools&lt;/strong&gt;, what the agent can do (&lt;code&gt;read&lt;/code&gt;, &lt;code&gt;write&lt;/code&gt;, &lt;code&gt;exec&lt;/code&gt;,&lt;br&gt;
&lt;code&gt;browser&lt;/code&gt;, &lt;code&gt;web_search&lt;/code&gt;; &lt;a href="https://docs.openclaw.ai/tools" rel="noopener noreferrer"&gt;tools&lt;/a&gt;); &lt;strong&gt;memory&lt;/strong&gt;,&lt;br&gt;
&lt;code&gt;MEMORY.md&lt;/code&gt; plus dated notes in &lt;code&gt;memory/&lt;/code&gt;, so it remembers only what it wrote down&lt;br&gt;
(&lt;a href="https://docs.openclaw.ai/concepts/memory" rel="noopener noreferrer"&gt;memory&lt;/a&gt;); &lt;strong&gt;cron&lt;/strong&gt; jobs, a prompt on a&lt;br&gt;
schedule delivered to a channel (&lt;a href="https://docs.openclaw.ai/cli/cron" rel="noopener noreferrer"&gt;automations&lt;/a&gt;); the&lt;br&gt;
&lt;strong&gt;heartbeat&lt;/strong&gt;, a check every 30 minutes by default, silent when nothing needs attention&lt;br&gt;
(&lt;a href="https://docs.openclaw.ai/gateway/heartbeat" rel="noopener noreferrer"&gt;heartbeat&lt;/a&gt;).&lt;/p&gt;
&lt;h2&gt;
  
  
  What it does, and what it does not
&lt;/h2&gt;

&lt;p&gt;It answers in your chat apps; reads and writes files in its workspace; runs commands on&lt;br&gt;
its machine; opens a browser; searches the web; remembers across days in files you can&lt;br&gt;
read; messages you first, on a schedule or when a heartbeat finds something; runs several&lt;br&gt;
agents and sub-agents. The docs' own examples: inbox and calendar briefings, research,&lt;br&gt;
reminders, browser automation, a task sent from your phone&lt;br&gt;
(&lt;a href="https://docs.openclaw.ai/help/faq/what-is-openclaw" rel="noopener noreferrer"&gt;FAQ&lt;/a&gt;).&lt;/p&gt;

&lt;p&gt;What it does not do, from the same sources:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;It has no model of its own.&lt;/strong&gt; You bring a provider key, a ChatGPT or Claude
subscription login, or a local model; the model is where the bill comes from. Every
path, including the free ones: &lt;a href="https://aplite.ai/blog/openclaw-without-api-key?utm_source=devto&amp;amp;utm_medium=article&amp;amp;utm_campaign=post&amp;amp;utm_content=what-is-openclaw" rel="noopener noreferrer"&gt;OpenClaw without an API key&lt;/a&gt;.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;It does not run while its machine is off.&lt;/strong&gt; Sleep, network drops and OS updates stop
the gateway; the upstream FAQ says to use a VPS after disconnects
(&lt;a href="https://docs.openclaw.ai/help/faq-first-run/providers-and-hosting" rel="noopener noreferrer"&gt;providers and hosting&lt;/a&gt;).
The free ways of keeping it on: &lt;a href="https://aplite.ai/blog/openclaw-hosting-for-free?utm_source=devto&amp;amp;utm_medium=article&amp;amp;utm_campaign=post&amp;amp;utm_content=what-is-openclaw" rel="noopener noreferrer"&gt;OpenClaw hosting for free&lt;/a&gt;.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;It is not locked down by default.&lt;/strong&gt; Sandboxing and exec approvals are off until you
turn them on, and one gateway is one trust domain: "a trusted single-operator
assistant" (&lt;a href="https://docs.openclaw.ai/start/why-openclaw" rel="noopener noreferrer"&gt;why OpenClaw&lt;/a&gt;,
&lt;a href="https://docs.openclaw.ai/gateway/security" rel="noopener noreferrer"&gt;security&lt;/a&gt;).&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;It is not a coding IDE.&lt;/strong&gt; The docs say Claude Code or Codex is better for fast coding
inside a repository; OpenClaw is the assistant around them, and can run both as its
agent runtime
(&lt;a href="https://docs.openclaw.ai/concepts/agent-runtimes" rel="noopener noreferrer"&gt;agent runtimes&lt;/a&gt;).&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;It is not a hosted service.&lt;/strong&gt; Hosting is what third parties sell,
&lt;a href="https://aplite.ai/blog/managed-openclaw-hosting-compared?utm_source=devto&amp;amp;utm_medium=article&amp;amp;utm_campaign=post&amp;amp;utm_content=what-is-openclaw" rel="noopener noreferrer"&gt;seventeen of them&lt;/a&gt; in October 2026; the
hosted path below is one.&lt;/li&gt;
&lt;/ul&gt;
&lt;h2&gt;
  
  
  How it differs from a chat with a model
&lt;/h2&gt;

&lt;p&gt;ChatGPT, Claude.ai and Gemini are a model behind a website: you go to it, the&lt;br&gt;
conversation lives there, and it touches what the vendor connected. OpenClaw inverts&lt;br&gt;
that: the assistant comes to the messenger you already have open, its memory is files on&lt;br&gt;
your disk, its hands are tools on your machine, and it acts on a schedule without being&lt;br&gt;
asked, which no chat tab does. The model is a setting: &lt;code&gt;/new &amp;lt;model&amp;gt;&lt;/code&gt; switches it&lt;br&gt;
mid-conversation, and an agent can fail over to a second provider. The price is that you&lt;br&gt;
operate it: machine, keys, updates and security are yours.&lt;/p&gt;

&lt;p&gt;Beside it in the same search: &lt;a href="https://github.com/NousResearch/hermes-agent" rel="noopener noreferrer"&gt;Hermes Agent&lt;/a&gt;&lt;br&gt;
by Nous Research, the same shape, self-hosted and messaging-first; and Claude Code and&lt;br&gt;
Codex, coding harnesses OpenClaw can drive rather than replace.&lt;/p&gt;
&lt;h2&gt;
  
  
  How to try it
&lt;/h2&gt;
&lt;h3&gt;
  
  
  Install it
&lt;/h3&gt;

&lt;p&gt;About five minutes, by the upstream&lt;br&gt;
&lt;a href="https://docs.openclaw.ai/start/getting-started" rel="noopener noreferrer"&gt;getting started&lt;/a&gt; page.&lt;/p&gt;

&lt;ol&gt;
&lt;li&gt;&lt;p&gt;&lt;strong&gt;Have Node 24.16 or newer, or 26.1 or newer&lt;/strong&gt; (the installer adds it when missing)&lt;br&gt;
and a model: a provider key, a Claude Code or Codex login, or Ollama.&lt;/p&gt;&lt;/li&gt;
&lt;li&gt;
&lt;p&gt;&lt;strong&gt;Install.&lt;/strong&gt; macOS, Linux or WSL2:&lt;br&gt;
&lt;/p&gt;

&lt;pre class="highlight plaintext"&gt;&lt;code&gt;curl -fsSL https://openclaw.ai/install.sh | bash
&lt;/code&gt;&lt;/pre&gt;



&lt;p&gt;Windows, in PowerShell:&lt;br&gt;
&lt;/p&gt;

&lt;pre class="highlight plaintext"&gt;&lt;code&gt;iwr -useb https://openclaw.ai/install.ps1 | iex
&lt;/code&gt;&lt;/pre&gt;



&lt;p&gt;Desktop apps: &lt;a href="https://docs.openclaw.ai/install" rel="noopener noreferrer"&gt;install page&lt;/a&gt;.&lt;/p&gt;
&lt;/li&gt;
&lt;li&gt;&lt;p&gt;&lt;strong&gt;Onboard.&lt;/strong&gt; The installer opens the wizard; &lt;em&gt;Quick start&lt;/em&gt; finds your login or key,&lt;br&gt;
verifies it with a real completion and opens the dashboard.&lt;/p&gt;&lt;/li&gt;
&lt;li&gt;
&lt;p&gt;&lt;strong&gt;Keep it running.&lt;/strong&gt; Press Ctrl+C in the foreground gateway, install it as a&lt;br&gt;
service, then check it:&lt;br&gt;
&lt;/p&gt;

&lt;pre class="highlight plaintext"&gt;&lt;code&gt;openclaw gateway install
openclaw gateway status
&lt;/code&gt;&lt;/pre&gt;



&lt;p&gt;The second command should show the gateway listening on port 18789.&lt;/p&gt;
&lt;/li&gt;
&lt;li&gt;&lt;p&gt;&lt;strong&gt;Talk to it.&lt;/strong&gt; &lt;code&gt;openclaw dashboard&lt;/code&gt; opens the Control UI; type a message. From your&lt;br&gt;
phone, Telegram is the fastest channel:&lt;br&gt;
&lt;a href="https://aplite.ai/blog/openclaw-telegram-setup?utm_source=devto&amp;amp;utm_medium=article&amp;amp;utm_campaign=post&amp;amp;utm_content=what-is-openclaw" rel="noopener noreferrer"&gt;OpenClaw in Telegram in five minutes&lt;/a&gt;.&lt;/p&gt;&lt;/li&gt;
&lt;/ol&gt;

&lt;p&gt;If something fails, &lt;code&gt;openclaw doctor&lt;/code&gt; reads the findings.&lt;/p&gt;

&lt;h3&gt;
  
  
  Without installing anything
&lt;/h3&gt;

&lt;p&gt;&lt;a href="https://aplite.ai/docs/openclaw?utm_source=devto&amp;amp;utm_medium=article&amp;amp;utm_campaign=post&amp;amp;utm_content=what-is-openclaw" rel="noopener noreferrer"&gt;Aplite&lt;/a&gt; runs a standard OpenClaw instance for you: one gateway in its own&lt;br&gt;
container on our server, currently 2026.9.8, set up from the browser. Create an account&lt;br&gt;
with email or Google, no card on the Free plan; pick an agent, press Deploy; about a&lt;br&gt;
minute later the workspace opens and you ask. No key to paste: models come through one&lt;br&gt;
metered account, billed in credits, 1 credit = 1¢, token usage under each reply. On top&lt;br&gt;
sit a markdown workspace the agent reads and writes, an approval queue for its changes,&lt;br&gt;
and Telegram from a bot token. What a managed instance gives up, the shell, raw&lt;br&gt;
&lt;code&gt;openclaw.json&lt;/code&gt;, local models and day-one releases, is on the&lt;br&gt;
&lt;a href="https://aplite.ai/docs/openclaw?utm_source=devto&amp;amp;utm_medium=article&amp;amp;utm_campaign=post&amp;amp;utm_content=what-is-openclaw" rel="noopener noreferrer"&gt;OpenClaw page&lt;/a&gt;; other hosts are in&lt;br&gt;
&lt;a href="https://aplite.ai/blog/managed-openclaw-hosting-compared?utm_source=devto&amp;amp;utm_medium=article&amp;amp;utm_campaign=post&amp;amp;utm_content=what-is-openclaw" rel="noopener noreferrer"&gt;Managed OpenClaw hosting compared&lt;/a&gt;. Aplite is&lt;br&gt;
independent and not affiliated with the OpenClaw project.&lt;/p&gt;

&lt;h2&gt;
  
  
  FAQ
&lt;/h2&gt;

&lt;h3&gt;
  
  
  What is OpenClaw?
&lt;/h3&gt;

&lt;p&gt;An open-source AI assistant you run on your own computer or server: it connects a model&lt;br&gt;
of your choice to your chat apps, gives it tools such as files, shell, browser and web&lt;br&gt;
search, keeps its memory in markdown files, and can act on a schedule.&lt;/p&gt;

&lt;h3&gt;
  
  
  Is OpenClaw free?
&lt;/h3&gt;

&lt;p&gt;The software is free and the project sells nothing. Running it costs a machine that stays&lt;br&gt;
on and the model, billed per token unless it runs locally&lt;br&gt;
(&lt;a href="https://aplite.ai/blog/openclaw-hosting-for-free?utm_source=devto&amp;amp;utm_medium=article&amp;amp;utm_campaign=post&amp;amp;utm_content=what-is-openclaw" rel="noopener noreferrer"&gt;OpenClaw hosting for free&lt;/a&gt;).&lt;/p&gt;

&lt;h3&gt;
  
  
  Do I need an API key to use OpenClaw?
&lt;/h3&gt;

&lt;p&gt;Self-hosted, yes, or a substitute: a ChatGPT or Claude subscription login that onboarding&lt;br&gt;
reuses, or a local model through Ollama. On a managed host that holds the model account,&lt;br&gt;
Aplite among them, there is no key.&lt;/p&gt;

&lt;h3&gt;
  
  
  How does OpenClaw work?
&lt;/h3&gt;

&lt;p&gt;The gateway stays running. A message from a chat app goes to an agent with its&lt;br&gt;
instructions, memory and skills; the agent calls the model, runs the tools the answer&lt;br&gt;
needs, and the reply returns through the same channel. Scheduled jobs and the heartbeat&lt;br&gt;
start the same loop without a message.&lt;/p&gt;

&lt;h3&gt;
  
  
  Who makes OpenClaw, and is it owned by OpenAI?
&lt;/h3&gt;

&lt;p&gt;Peter Steinberger published it in late 2025 as Clawdbot, renamed Moltbot and then&lt;br&gt;
OpenClaw in January 2026 (&lt;a href="https://en.wikipedia.org/wiki/OpenClaw" rel="noopener noreferrer"&gt;Wikipedia&lt;/a&gt;); the&lt;br&gt;
OpenClaw Foundation, an independent non-profit funded by donations, stewards it since he&lt;br&gt;
joined OpenAI in February 2026. OpenAI is one donor; the docs say donors do not control&lt;br&gt;
the project and no lab's model is privileged in the code.&lt;/p&gt;

</description>
      <category>openclaw</category>
      <category>ai</category>
      <category>opensource</category>
      <category>beginners</category>
    </item>
    <item>
      <title>OpenClaw hosting for free: what each option actually gives you</title>
      <dc:creator>Aplite</dc:creator>
      <pubDate>Fri, 16 Oct 2026 06:50:59 +0000</pubDate>
      <link>https://dev.to/aplite/openclaw-hosting-for-free-what-each-option-actually-gives-you-2g1c</link>
      <guid>https://dev.to/aplite/openclaw-hosting-for-free-what-each-option-actually-gives-you-2g1c</guid>
      <description>&lt;p&gt;&lt;em&gt;I work on &lt;a href="https://aplite.ai/docs/openclaw?utm_source=devto&amp;amp;utm_medium=article&amp;amp;utm_campaign=post&amp;amp;utm_content=openclaw-hosting-for-free" rel="noopener noreferrer"&gt;Aplite&lt;/a&gt;, a managed OpenClaw host. It is one of the options compared below; the prices and terms of the others are from their own pages, linked in the table.&lt;/em&gt;&lt;/p&gt;

&lt;p&gt;Search for &lt;em&gt;openclaw hosting free&lt;/em&gt; and the first page is three kinds of result: hosts whose&lt;br&gt;
"free" is a trial or a first-month price, Oracle free-tier guides (several still quoting an&lt;br&gt;
Arm allowance Oracle has since halved), and "best hosting" lists written by the hosts. What&lt;br&gt;
none of them does is the accounting. An agent that works around the clock costs two things —&lt;br&gt;
a machine that stays on, and model tokens for every reply — and most free offers cover one&lt;br&gt;
of the two, for a while.&lt;/p&gt;

&lt;p&gt;Here is that accounting as of October 2026, and the one path that costs nothing, with every&lt;br&gt;
command run on OpenClaw 2026.9.8. If you would rather not run a server at all, the same&lt;br&gt;
runtime is hosted in three steps further down.&lt;/p&gt;
&lt;h2&gt;
  
  
  What "free" has to cover
&lt;/h2&gt;

&lt;p&gt;&lt;strong&gt;The server.&lt;/strong&gt; The gateway is a Node process that has to stay up. The upstream&lt;br&gt;
&lt;a href="https://docs.openclaw.ai/help/faq-first-run/providers-and-hosting" rel="noopener noreferrer"&gt;hardware FAQ&lt;/a&gt; puts the&lt;br&gt;
floor at 1 vCPU, 1 GB of RAM and about 500 MB of disk, and recommends 2 GB or more. The floor&lt;br&gt;
is real: capped at 1 GB, the 2026.9.8 gateway settles at about 740 MB of resident memory a&lt;br&gt;
minute after start and logs &lt;code&gt;memory pressure: level=warning&lt;/code&gt;. The smallest free cloud&lt;br&gt;
machines give you exactly 1 GB.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;The model.&lt;/strong&gt; Every reply is a model call, billed per token wherever the gateway runs.&lt;br&gt;
Three free ways: a local model through &lt;a href="https://docs.openclaw.ai/providers/ollama" rel="noopener noreferrer"&gt;Ollama&lt;/a&gt;&lt;br&gt;
(needs a capable machine; the same FAQ warns that small quantized models truncate context&lt;br&gt;
and raise prompt-injection risk); OpenRouter's &lt;code&gt;:free&lt;/code&gt; variants, limited to&lt;br&gt;
&lt;a href="https://openrouter.ai/docs/api-reference/limits" rel="noopener noreferrer"&gt;20 requests a minute and 50 a day&lt;/a&gt; until&lt;br&gt;
you have bought $10 of credits, then 1,000 a day; or the&lt;br&gt;
&lt;a href="https://ai.google.dev/gemini-api/docs/pricing" rel="noopener noreferrer"&gt;Gemini API free tier&lt;/a&gt;, where Flash is free&lt;br&gt;
of charge and, in Google's words, "content used to improve our products". Hosted plans that&lt;br&gt;
include credits cover tokens for as long as the credits last.&lt;/p&gt;
&lt;h2&gt;
  
  
  The options, side by side
&lt;/h2&gt;

&lt;p&gt;Prices and terms as each page read on 13 October 2026; the links go to those pages.&lt;/p&gt;

&lt;div class="table-wrapper-paragraph"&gt;&lt;table&gt;
&lt;thead&gt;
&lt;tr&gt;
&lt;th&gt;Option&lt;/th&gt;
&lt;th&gt;Price, October 2026&lt;/th&gt;
&lt;th&gt;You set up&lt;/th&gt;
&lt;th&gt;Stops being free&lt;/th&gt;
&lt;th&gt;Watch for&lt;/th&gt;
&lt;/tr&gt;
&lt;/thead&gt;
&lt;tbody&gt;
&lt;tr&gt;
&lt;td&gt;Your own machine&lt;/td&gt;
&lt;td&gt;$0&lt;/td&gt;
&lt;td&gt;install, a model, keeping it awake&lt;/td&gt;
&lt;td&gt;never&lt;/td&gt;
&lt;td&gt;sleep, updates and network drops stop the gateway&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;&lt;a href="https://docs.oracle.com/en-us/iaas/Content/FreeTier/freetier.htm" rel="noopener noreferrer"&gt;Oracle Cloud Always Free&lt;/a&gt;&lt;/td&gt;
&lt;td&gt;$0; card and phone number at sign-up, not charged&lt;/td&gt;
&lt;td&gt;VM, firewall, install&lt;/td&gt;
&lt;td&gt;no end date&lt;/td&gt;
&lt;td&gt;the Arm allowance is &lt;a href="https://docs.oracle.com/en-us/iaas/Content/FreeTier/freetier_topic-Always_Free_Resources.htm" rel="noopener noreferrer"&gt;2 OCPUs and 12 GB&lt;/a&gt;, not the 4 and 24 older guides quote; "Out of host capacity" on creation; instances idle for 7 days can be reclaimed&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;&lt;a href="https://docs.cloud.google.com/free/docs/free-cloud-features" rel="noopener noreferrer"&gt;Google Cloud free tier&lt;/a&gt;&lt;/td&gt;
&lt;td&gt;$0 for one e2-micro in three US regions, 30 GB disk; card required&lt;/td&gt;
&lt;td&gt;VM, firewall, install&lt;/td&gt;
&lt;td&gt;no end date; the $300 trial credit lasts 90 days&lt;/td&gt;
&lt;td&gt;1 GB of memory — the floor&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;&lt;a href="https://aws.amazon.com/free/" rel="noopener noreferrer"&gt;AWS Free Tier&lt;/a&gt;&lt;/td&gt;
&lt;td&gt;$100 in credits at sign-up, up to $200&lt;/td&gt;
&lt;td&gt;VM, firewall, install&lt;/td&gt;
&lt;td&gt;after 6 months, or when the credits run out&lt;/td&gt;
&lt;td&gt;the old always-free micro instance is gone&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Cheapest VPS (&lt;a href="https://www.hostinger.com/vps-hosting" rel="noopener noreferrer"&gt;Hostinger KVM 1&lt;/a&gt;)&lt;/td&gt;
&lt;td&gt;$6.49/mo on a 24-month term, renews at $11.99; 1 vCPU, 4 GB&lt;/td&gt;
&lt;td&gt;install&lt;/td&gt;
&lt;td&gt;never free; 30-day money-back&lt;/td&gt;
&lt;td&gt;the renewal price&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;&lt;a href="https://runware.ai/free-openclaw" rel="noopener noreferrer"&gt;Runware&lt;/a&gt;&lt;/td&gt;
&lt;td&gt;hosting $0, "AI inference billed per token"; $2 in credit with a business email, no payment to deploy&lt;/td&gt;
&lt;td&gt;nothing&lt;/td&gt;
&lt;td&gt;hosting stays free; tokens are paid from the first reply&lt;/td&gt;
&lt;td&gt;one instance per account, &lt;a href="https://dev.to/runware/we-hosted-openclaw-so-you-dont-have-to-3698"&gt;suspended after 5 days unused&lt;/a&gt;
&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;KiloClaw&lt;/td&gt;
&lt;td&gt;no price on kilo.ai: the product pages redirect to the homepage, the pricing page does not mention it&lt;/td&gt;
&lt;td&gt;—&lt;/td&gt;
&lt;td&gt;—&lt;/td&gt;
&lt;td&gt;
&lt;a href="https://vibecoding.app/blog/kiloclaw-review" rel="noopener noreferrer"&gt;reviews&lt;/a&gt; report $55/mo with a one-day trial since August 2026, $9/mo with 7 days before that&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;&lt;a href="https://www.hostinger.com/vps/openclaw-hosting" rel="noopener noreferrer"&gt;Hostinger managed OpenClaw&lt;/a&gt;&lt;/td&gt;
&lt;td&gt;$5.99/mo on a 24-month term, $10.99 month to month, renews at $11.99; AI credits included&lt;/td&gt;
&lt;td&gt;nothing&lt;/td&gt;
&lt;td&gt;never free; 30-day money-back&lt;/td&gt;
&lt;td&gt;no trial&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;&lt;a href="https://xcloud.host/openclaw-hosting/" rel="noopener noreferrer"&gt;xCloud&lt;/a&gt;&lt;/td&gt;
&lt;td&gt;$9.99 the first month for 6 GB, then $19.99; your own API key&lt;/td&gt;
&lt;td&gt;nothing&lt;/td&gt;
&lt;td&gt;never free&lt;/td&gt;
&lt;td&gt;the renewal price&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;&lt;a href="https://www.ampere.sh/" rel="noopener noreferrer"&gt;Ampere&lt;/a&gt;&lt;/td&gt;
&lt;td&gt;the "Free" tier is listed at $1/mo; the FAQ says free hosting during beta, $10 in AI credits, no card&lt;/td&gt;
&lt;td&gt;nothing&lt;/td&gt;
&lt;td&gt;after the beta&lt;/td&gt;
&lt;td&gt;the plan table and the FAQ disagree&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;&lt;a href="https://rapidclaw.dev/" rel="noopener noreferrer"&gt;RapidClaw&lt;/a&gt;&lt;/td&gt;
&lt;td&gt;5 free messages, then $29/mo; no card to start&lt;/td&gt;
&lt;td&gt;nothing&lt;/td&gt;
&lt;td&gt;after 5 messages&lt;/td&gt;
&lt;td&gt;—&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;&lt;a href="https://aplite.ai/docs/openclaw?utm_source=devto&amp;amp;utm_medium=article&amp;amp;utm_campaign=post&amp;amp;utm_content=openclaw-hosting-for-free" rel="noopener noreferrer"&gt;Aplite Free&lt;/a&gt;&lt;/td&gt;
&lt;td&gt;$0, no card&lt;/td&gt;
&lt;td&gt;nothing — set up from the browser&lt;/td&gt;
&lt;td&gt;no end date; credit packs and schedules are on Pro ($20/mo)&lt;/td&gt;
&lt;td&gt;one agent, budget models, no scheduled automations&lt;/td&gt;
&lt;/tr&gt;
&lt;/tbody&gt;
&lt;/table&gt;&lt;/div&gt;
&lt;h2&gt;
  
  
  The $0 path, step by step
&lt;/h2&gt;

&lt;p&gt;Run in a fresh container of our 2026.9.8 runtime image with a 1 GB memory cap, and checked&lt;br&gt;
against the &lt;a href="https://docs.openclaw.ai/start/getting-started" rel="noopener noreferrer"&gt;upstream docs&lt;/a&gt;. Your install will&lt;br&gt;
be newer — the gateway says so the moment it starts — but the commands are the same.&lt;/p&gt;

&lt;ol&gt;
&lt;li&gt;
&lt;p&gt;&lt;strong&gt;Install.&lt;/strong&gt; OpenClaw needs Node 24.16 or 26.1 and newer&lt;br&gt;
(&lt;a href="https://docs.openclaw.ai/install" rel="noopener noreferrer"&gt;install docs&lt;/a&gt;); the installer provisions it when&lt;br&gt;
missing. On macOS, Linux or WSL2:&lt;br&gt;
&lt;/p&gt;

&lt;pre class="highlight plaintext"&gt;&lt;code&gt;curl -fsSL https://openclaw.ai/install.sh | bash
&lt;/code&gt;&lt;/pre&gt;



&lt;p&gt;On Windows, in PowerShell:&lt;br&gt;
&lt;/p&gt;

&lt;pre class="highlight plaintext"&gt;&lt;code&gt;iwr -useb https://openclaw.ai/install.ps1 | iex
&lt;/code&gt;&lt;/pre&gt;



&lt;p&gt;Then &lt;code&gt;openclaw --version&lt;/code&gt; — ours printed &lt;code&gt;OpenClaw 2026.9.8 (fc23bc8)&lt;/code&gt;.&lt;/p&gt;
&lt;/li&gt;
&lt;li&gt;
&lt;p&gt;&lt;strong&gt;Have a free model ready.&lt;/strong&gt; Local: install Ollama and pull a model&lt;br&gt;
(&lt;code&gt;ollama pull gemma4&lt;/code&gt; in the &lt;a href="https://docs.openclaw.ai/providers/ollama" rel="noopener noreferrer"&gt;upstream recipe&lt;/a&gt;).&lt;br&gt;
Or get a key from OpenRouter or Google AI Studio. Pick Ollama with the server not running&lt;br&gt;
and onboarding stops here:&lt;br&gt;
&lt;/p&gt;

&lt;pre class="highlight plaintext"&gt;&lt;code&gt;Ollama could not be reached at http://127.0.0.1:11434.
Start or restart the Ollama server for this address.
If Ollama is not installed on that machine, download it at https://ollama.com/download
&lt;/code&gt;&lt;/pre&gt;


&lt;/li&gt;
&lt;li&gt;
&lt;p&gt;&lt;strong&gt;Run onboarding.&lt;/strong&gt;&lt;br&gt;
&lt;/p&gt;

&lt;pre class="highlight plaintext"&gt;&lt;code&gt;openclaw onboard
&lt;/code&gt;&lt;/pre&gt;



&lt;p&gt;Choose &lt;em&gt;Quick start&lt;/em&gt;, pick the provider, paste the key. OpenClaw verifies it with a&lt;br&gt;
real completion, writes &lt;code&gt;~/.openclaw/openclaw.json&lt;/code&gt;, starts the gateway in the&lt;br&gt;
foreground and opens the dashboard. Skip the provider and everything still installs,&lt;br&gt;
but the first message fails with:&lt;br&gt;
&lt;/p&gt;

&lt;pre class="highlight plaintext"&gt;&lt;code&gt;No API key found for provider "openrouter".
Auth store: ~/.openclaw/agents/main/agent/openclaw-agent.sqlite
Configure an API key (openclaw models auth paste-api-key --provider openrouter)
| missing-provider-auth
&lt;/code&gt;&lt;/pre&gt;



&lt;p&gt;That command is the fix: &lt;code&gt;openclaw models auth paste-api-key --provider openrouter&lt;/code&gt;.&lt;/p&gt;
&lt;/li&gt;
&lt;li&gt;
&lt;p&gt;&lt;strong&gt;Check the gateway.&lt;/strong&gt; In a second terminal, &lt;code&gt;openclaw gateway status&lt;/code&gt; prints:&lt;br&gt;
&lt;/p&gt;

&lt;pre class="highlight plaintext"&gt;&lt;code&gt;Gateway: bind=loopback (127.0.0.1), port=18789 (env/config)
Dashboard: http://127.0.0.1:18789/
Connectivity probe: ok
Listening: 127.0.0.1:18789
&lt;/code&gt;&lt;/pre&gt;



&lt;p&gt;&lt;code&gt;openclaw dashboard&lt;/code&gt; opens the browser with a one-time pairing token in the URL. With&lt;br&gt;
no provider set, the Control UI lands on &lt;em&gt;Model Setup&lt;/em&gt;.&lt;/p&gt;
&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/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2F7wdcij6c0bvls01j5nfz.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%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2F7wdcij6c0bvls01j5nfz.webp" alt="The Control UI on a fresh install: the gateway is up, no model is connected yet." width="800" height="500"&gt;&lt;/a&gt;&lt;br&gt;
&lt;em&gt;The Control UI on a fresh install: the gateway is up, no model is connected yet.&lt;/em&gt;&lt;/p&gt;

&lt;ol&gt;
&lt;li&gt;
&lt;p&gt;&lt;strong&gt;Keep it running after you log out.&lt;/strong&gt;&lt;br&gt;
&lt;/p&gt;

&lt;pre class="highlight plaintext"&gt;&lt;code&gt;openclaw gateway install
&lt;/code&gt;&lt;/pre&gt;



&lt;p&gt;This installs a systemd user unit on Linux and WSL2, a LaunchAgent on macOS, a&lt;br&gt;
Scheduled Task on Windows (&lt;a href="https://docs.openclaw.ai/gateway" rel="noopener noreferrer"&gt;gateway docs&lt;/a&gt;). On a&lt;br&gt;
server, add &lt;code&gt;sudo loginctl enable-linger $USER&lt;/code&gt; so the unit survives your SSH session,&lt;br&gt;
as the upstream &lt;a href="https://docs.openclaw.ai/install/oracle" rel="noopener noreferrer"&gt;Oracle guide&lt;/a&gt; does.&lt;/p&gt;
&lt;/li&gt;
&lt;/ol&gt;

&lt;p&gt;Where this path ends is in the same upstream FAQ: on a laptop, "sleep, network drops, and OS&lt;br&gt;
updates can interrupt the Gateway", and "use a VPS if you have had gateway disconnects&lt;br&gt;
before". On the free clouds it is the 1 GB floor and Oracle's reclaim of idle instances; on&lt;br&gt;
free models it is the request cap, since one agent turn with tools makes several model&lt;br&gt;
requests. That is when people pay — for a $6–12 VPS plus a key, or for a managed host.&lt;/p&gt;

&lt;h2&gt;
  
  
  Hosted: the same runtime in three steps
&lt;/h2&gt;

&lt;p&gt;&lt;a href="https://aplite.ai/docs/openclaw?utm_source=devto&amp;amp;utm_medium=article&amp;amp;utm_campaign=post&amp;amp;utm_content=openclaw-hosting-for-free" rel="noopener noreferrer"&gt;Aplite&lt;/a&gt; runs a standard OpenClaw instance for you, set up from the&lt;br&gt;
browser. The whole path:&lt;/p&gt;

&lt;ol&gt;
&lt;li&gt;&lt;p&gt;&lt;strong&gt;Create an account.&lt;/strong&gt; Email or Google, no card. The &lt;a href="https://aplite.ai/docs/plans?utm_source=devto&amp;amp;utm_medium=article&amp;amp;utm_campaign=post&amp;amp;utm_content=openclaw-hosting-for-free" rel="noopener noreferrer"&gt;Free plan&lt;/a&gt; is one&lt;br&gt;
agent, the full workspace, budget models and 300 welcome credits that never expire.&lt;/p&gt;&lt;/li&gt;
&lt;li&gt;&lt;p&gt;&lt;strong&gt;Deploy.&lt;/strong&gt; Pick the agents and a model for each, press Deploy. About a minute later the&lt;br&gt;
workspace opens with its own OpenClaw instance — currently 2026.9.8 — in its own&lt;br&gt;
container.&lt;/p&gt;&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/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fdexbhbyctos1tpo0f9tu.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%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fdexbhbyctos1tpo0f9tu.webp" alt="A Free workspace: a standard OpenClaw runtime in its own container." width="800" height="500"&gt;&lt;/a&gt;&lt;br&gt;
&lt;em&gt;A Free workspace: a standard OpenClaw runtime in its own container.&lt;/em&gt;&lt;/p&gt;

&lt;ol&gt;
&lt;li&gt;
&lt;strong&gt;Ask.&lt;/strong&gt; Nothing to install and no key to paste: models come through one metered
account, 1 credit is 1¢, spent only when the agent calls a model, and the cost sits
under every reply.&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/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fg3r0uo4yjex0veqxhc22.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%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fg3r0uo4yjex0veqxhc22.webp" alt="The agent answers in the browser with no key pasted anywhere; the line under the reply is what it cost." width="800" height="500"&gt;&lt;/a&gt;&lt;br&gt;
&lt;em&gt;The agent answers in the browser with no key pasted anywhere; the line under the reply is what it cost.&lt;/em&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%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fatzimsyednmqpi9s3bal.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%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fatzimsyednmqpi9s3bal.webp" alt="Settings → Cost: the welcome credits, and what the week spent." width="800" height="500"&gt;&lt;/a&gt;&lt;br&gt;
&lt;em&gt;Settings → Cost: the welcome credits, and what the week spent.&lt;/em&gt;&lt;/p&gt;

&lt;p&gt;On top of the runtime sit a markdown workspace the agent reads and writes, an approval&lt;br&gt;
queue, and Telegram delivery from a bot token; scheduled routines are on Pro ($20/mo). What&lt;br&gt;
a managed instance gives up — the shell, raw &lt;code&gt;openclaw.json&lt;/code&gt;, local models, day-one releases&lt;br&gt;
— is listed on the &lt;a href="https://aplite.ai/docs/openclaw?utm_source=devto&amp;amp;utm_medium=article&amp;amp;utm_campaign=post&amp;amp;utm_content=openclaw-hosting-for-free" rel="noopener noreferrer"&gt;OpenClaw page&lt;/a&gt;. Aplite is an independent service and is&lt;br&gt;
not affiliated with the OpenClaw project.&lt;/p&gt;

&lt;h2&gt;
  
  
  FAQ
&lt;/h2&gt;

&lt;h3&gt;
  
  
  Is OpenClaw itself free?
&lt;/h3&gt;

&lt;p&gt;Yes: open-source software under the MIT license, and the&lt;br&gt;
&lt;a href="https://docs.openclaw.ai/install" rel="noopener noreferrer"&gt;install&lt;/a&gt; costs nothing. What costs money is the machine it&lt;br&gt;
runs on and the model it calls, which is what "free hosting" offers differ on.&lt;/p&gt;

&lt;h3&gt;
  
  
  Can I host OpenClaw for free forever?
&lt;/h3&gt;

&lt;p&gt;On your own machine with a local model through Ollama, yes — as long as the machine stays&lt;br&gt;
on. Oracle's Always Free tier has no end date but reclaims idle instances and is often out of&lt;br&gt;
capacity; Runware charges nothing for hosting and bills tokens; Aplite's Free plan has no end&lt;br&gt;
date and welcome credits that do not expire. The rest are trials or credits.&lt;/p&gt;

&lt;h3&gt;
  
  
  Do I need a credit card for free OpenClaw hosting?
&lt;/h3&gt;

&lt;p&gt;Not for your own machine, Runware, RapidClaw, Ampere during its beta, or Aplite Free. Oracle&lt;br&gt;
takes a card and a phone number at sign-up without charging; Google Cloud requires a payment&lt;br&gt;
method; Hostinger and xCloud are paid from day one.&lt;/p&gt;

&lt;h3&gt;
  
  
  Do I need an API key?
&lt;/h3&gt;

&lt;p&gt;Self-hosting, yes — unless the model runs locally with Ollama; the key is where the token&lt;br&gt;
bill goes. On Aplite there is no key: models run through a metered account and usage is&lt;br&gt;
billed in credits.&lt;/p&gt;




&lt;p&gt;&lt;em&gt;Originally published on &lt;a href="https://aplite.ai/blog/openclaw-hosting-for-free" rel="noopener noreferrer"&gt;aplite.ai&lt;/a&gt;. Try it free at &lt;a href="https://aplite.ai/?utm_source=devto&amp;amp;utm_medium=article&amp;amp;utm_campaign=post&amp;amp;utm_content=openclaw-hosting-for-free" rel="noopener noreferrer"&gt;aplite.ai&lt;/a&gt;.&lt;/em&gt;&lt;/p&gt;

</description>
      <category>openclaw</category>
      <category>hosting</category>
      <category>selfhosting</category>
      <category>ai</category>
    </item>
  </channel>
</rss>
