<?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: Georges Petrov</title>
    <description>The latest articles on DEV Community by Georges Petrov (@gmpetrov).</description>
    <link>https://dev.to/gmpetrov</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%2F241042%2Fe488efbe-37f2-4cc0-90f7-1627fa8caccb.jpeg</url>
      <title>DEV Community: Georges Petrov</title>
      <link>https://dev.to/gmpetrov</link>
    </image>
    <atom:link rel="self" type="application/rss+xml" href="https://dev.to/feed/gmpetrov"/>
    <language>en</language>
    <item>
      <title>How I run support, marketing and code with a team of Claude Code and Codex agents</title>
      <dc:creator>Georges Petrov</dc:creator>
      <pubDate>Wed, 11 Nov 2026 15:23:50 +0000</pubDate>
      <link>https://dev.to/gmpetrov/how-i-run-support-marketing-and-code-with-a-team-of-claude-code-and-codex-agents-jp1</link>
      <guid>https://dev.to/gmpetrov/how-i-run-support-marketing-and-code-with-a-team-of-claude-code-and-codex-agents-jp1</guid>
      <description>&lt;p&gt;I run several small products on my own. For a while my setup was ten terminal tabs of Claude Code, each one halfway through something, and I could never tell which one was waiting on me. The coding worked. Everything around it (support email, follow-ups, the weekly numbers, posts) still waited until I remembered it.&lt;/p&gt;

&lt;p&gt;So I built Brigade (brigade.lol), an open-source (MIT) app that runs a team of AI teammates. Each teammate is the real Claude Code or Codex CLI, signed in with my own Claude or ChatGPT account, with a job, its own computer, the apps I give it and a shared memory of how we work.&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;Code: &lt;a href="https://github.com/gmpetrov/brigade" rel="noopener noreferrer"&gt;https://github.com/gmpetrov/brigade&lt;/a&gt;
&lt;/li&gt;
&lt;li&gt;Live demo, no sign-up: &lt;a href="https://demo.brigade.lol" rel="noopener noreferrer"&gt;https://demo.brigade.lol&lt;/a&gt;
&lt;/li&gt;
&lt;/ul&gt;

&lt;h2&gt;
  
  
  Why the official CLIs and not an API
&lt;/h2&gt;

&lt;p&gt;I already pay for Claude Max and ChatGPT Pro. I didn't want a second bill for tokens, and I didn't want to rebuild an agent loop that Anthropic and OpenAI improve every few weeks.&lt;/p&gt;

&lt;p&gt;So a small runner starts the official &lt;code&gt;claude&lt;/code&gt; or &lt;code&gt;codex&lt;/code&gt; command on a computer, signed in with the member's account. It strips API keys from its environment, so the work runs on the subscription. When one account hits its limit, the teammate moves to the next account of the same provider and carries on in the middle of the task.&lt;/p&gt;

&lt;p&gt;The runner is the only part that needs a computer: your own Mac or Linux box, or a cloud computer we host.&lt;/p&gt;

&lt;h2&gt;
  
  
  One job per teammate
&lt;/h2&gt;

&lt;p&gt;My first attempt was one agent with every tool. It got vague fast, and when something went wrong I couldn't tell which instruction caused it. Now each teammate holds one job. Support answers email from the help docs. Marketing drafts posts and keeps directory listings up to date. An engineer opens pull requests, starts a live preview of the app in the thread, and asks another teammate to review before I merge.&lt;/p&gt;

&lt;p&gt;Each one gets its own apps (Gmail, GitHub, Stripe, Notion, Linear, PostHog and others, or any MCP server), read-only or read-write, teammate by teammate. When one is off, I fix that one and leave the others alone.&lt;/p&gt;

&lt;p&gt;I talk to them in threads, the way I would with a new hire. What we decide goes into a workspace memory and a shared library of documents, so I no longer re-explain the product, the tone and last week's decision at the start of every session.&lt;/p&gt;

&lt;h2&gt;
  
  
  Work that starts without me
&lt;/h2&gt;

&lt;p&gt;A trigger starts a thread when something happens in an app or a webhook arrives: a failed payment, a new GitHub issue, an email. Routines run on a schedule, like the weekly report. Every step a teammate takes is logged in its thread, and there are daily caps on threads, computer hours and writes to each app.&lt;/p&gt;

&lt;h2&gt;
  
  
  What still comes to me
&lt;/h2&gt;

&lt;p&gt;For each teammate, I choose which app actions wait for my yes: sending an email, a change in Stripe, a merge. Those requests land in one list, "Needs you", with what the teammate wants to do and why. My phone gets a push, and a tap opens the request.&lt;/p&gt;

&lt;p&gt;This is opt-in. A teammate asks first only for the app actions I set it to ask about, and does the rest on its own. I started strict and loosened it teammate by teammate.&lt;/p&gt;

&lt;h2&gt;
  
  
  It builds itself
&lt;/h2&gt;

&lt;p&gt;Most of the commits in the repo come from the app's own teammates, on branches named &lt;code&gt;brigade/…&lt;/code&gt;. I make the calls that matter: what to build, what to merge, what to throw away. The merged pull requests are public if you want to see what that looks like day to day.&lt;/p&gt;

&lt;h2&gt;
  
  
  What doesn't work yet
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;On the free plan, teammates work only while your computer is on. Close the laptop and they stop. The paid plan ($99 a month per workspace) gives them a computer that keeps going.&lt;/li&gt;
&lt;li&gt;It runs Claude Code and Codex only. If you want Gemini, an open model or a local one, it can't run them today.&lt;/li&gt;
&lt;li&gt;Teaching a task (you do it once while the teammate watches) and taking over a teammate's screen work only on the cloud computer.&lt;/li&gt;
&lt;li&gt;Full computer use, where a teammate reads the screen and clicks in any app, is Mac only.&lt;/li&gt;
&lt;li&gt;Approvals cover the app actions you pick, nothing else. A teammate with read-write GitHub access and no approval set acts on its own.&lt;/li&gt;
&lt;li&gt;The name is shared. Other open-source projects are called Brigade (spinabot/brigade, escoffier-labs/brigade) and have nothing to do with this one, which is Brigade (brigade.lol).&lt;/li&gt;
&lt;/ul&gt;

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

&lt;p&gt;I wrote up how Brigade (brigade.lol) compares with OpenClaw, Paperclip, Lindy, Sintra, Polsia and plain Claude Code or Codex, including where it is weaker: &lt;a href="https://www.brigade.lol/compare" rel="noopener noreferrer"&gt;https://www.brigade.lol/compare&lt;/a&gt;&lt;/p&gt;

&lt;h2&gt;
  
  
  Try it
&lt;/h2&gt;

&lt;p&gt;It's free on your own computer (macOS or Linux), and the whole product is in the repo.&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;Live demo, no sign-up: &lt;a href="https://demo.brigade.lol" rel="noopener noreferrer"&gt;https://demo.brigade.lol&lt;/a&gt;
&lt;/li&gt;
&lt;li&gt;Source: &lt;a href="https://github.com/gmpetrov/brigade" rel="noopener noreferrer"&gt;https://github.com/gmpetrov/brigade&lt;/a&gt;
&lt;/li&gt;
&lt;li&gt;Hosted version: &lt;a href="https://www.brigade.lol" rel="noopener noreferrer"&gt;https://www.brigade.lol&lt;/a&gt;
&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;I'd like to hear where it breaks for you, here in the comments or in an issue.&lt;/p&gt;

</description>
      <category>showdev</category>
      <category>opensource</category>
      <category>ai</category>
      <category>claudecode</category>
    </item>
    <item>
      <title>I built an open-source tool that lets your coding agent film screen recordings</title>
      <dc:creator>Georges Petrov</dc:creator>
      <pubDate>Tue, 03 Nov 2026 11:28:58 +0000</pubDate>
      <link>https://dev.to/gmpetrov/i-built-an-open-source-tool-that-lets-your-coding-agent-film-screen-recordings-2k7n</link>
      <guid>https://dev.to/gmpetrov/i-built-an-open-source-tool-that-lets-your-coding-agent-film-screen-recordings-2k7n</guid>
      <description>&lt;p&gt;I like Screen Studio-style videos: a smooth cursor, zooms that follow the action, a nice gradient behind the window. Making them by hand means recording, re-recording, then editing. So I built &lt;strong&gt;Screenplay&lt;/strong&gt;, an open-source tool that lets your coding agent do it.&lt;/p&gt;

&lt;p&gt;Repo (MIT): &lt;a href="https://github.com/gmpetrov/agent-screenplay" rel="noopener noreferrer"&gt;https://github.com/gmpetrov/agent-screenplay&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%2Fabefts40r3ic4kxtbpjg.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%2Fabefts40r3ic4kxtbpjg.png" alt="Storyboard of a Screenplay recording" width="800" height="293"&gt;&lt;/a&gt;&lt;/p&gt;

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

&lt;p&gt;You ask your agent (Claude Code, Codex, Cursor or any MCP client) for a video:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight plaintext"&gt;&lt;code&gt;/screenplay Make a 30-second walkthrough of https://yoursite.com
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;ol&gt;
&lt;li&gt;The agent opens the site, explores it, and writes a short JSON &lt;strong&gt;scenario&lt;/strong&gt;: open this page, click this button, type that, hold here, zoom there.&lt;/li&gt;
&lt;li&gt;Screenplay runs &lt;code&gt;check&lt;/code&gt;, which plays the steps fast without filming, so broken selectors show up before the real take.&lt;/li&gt;
&lt;li&gt;
&lt;code&gt;record&lt;/code&gt; plays the scenario in headless Chromium (Playwright), films it, then edits it: smoothed cursor, automatic zooms, click ripples, callouts, spotlight, captions and an optional voice-over.&lt;/li&gt;
&lt;li&gt;The agent looks at the storyboard image of the take and fixes the scenario if a shot is off. Re-editing the look doesn't need a new take: &lt;code&gt;render&lt;/code&gt;.&lt;/li&gt;
&lt;/ol&gt;

&lt;p&gt;Desktop apps work too (macOS, Windows, Linux): the same kind of scenario is played through the OS accessibility APIs, and the window is filmed with ffmpeg.&lt;/p&gt;

&lt;h2&gt;
  
  
  A scenario
&lt;/h2&gt;

&lt;p&gt;Targets are named the way people see them (&lt;code&gt;text=…&lt;/code&gt;, &lt;code&gt;role=button[name='…']&lt;/code&gt;, &lt;code&gt;label=…&lt;/code&gt;), so the agent doesn't need CSS selectors:&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;"url"&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="s2"&gt;"https://app.example.com/inbox"&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt;
  &lt;span class="nl"&gt;"style"&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="p"&gt;{&lt;/span&gt; &lt;span class="nl"&gt;"background"&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="s2"&gt;"dawn"&lt;/span&gt; &lt;span class="p"&gt;}&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt;
  &lt;span class="nl"&gt;"steps"&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="nl"&gt;"click"&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="s2"&gt;"text=Camille Laurent"&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="nl"&gt;"say"&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="s2"&gt;"When the AI agent gets stuck, the conversation lands in your inbox."&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="nl"&gt;"callout"&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="s2"&gt;"role=button[name='Take over']"&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="nl"&gt;"text"&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="s2"&gt;"Take over from the AI agent"&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="nl"&gt;"click"&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="s2"&gt;"role=button[name='Take over']"&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="nl"&gt;"say"&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="s2"&gt;"One click, and the conversation is yours."&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="nl"&gt;"type"&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="s2"&gt;"Hi Camille, I'm on it."&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="nl"&gt;"into"&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="s2"&gt;"placeholder=Write a reply…"&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="nl"&gt;"click"&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="s2"&gt;"role=button[name='Send']"&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;h2&gt;
  
  
  Design choices
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;No AI inside, no API key.&lt;/strong&gt; Your agent does the thinking; Screenplay only plays, films and edits. It's a CLI plus an MCP server.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Headless first.&lt;/strong&gt; Coding agents often run in containers without a screen, and they still need to film.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Own renderer.&lt;/strong&gt; Frames are drawn with &lt;code&gt;@napi-rs/canvas&lt;/code&gt; and encoded with ffmpeg (both ship with the package). I didn't want a rendering library whose licence makes companies pay, in an MIT tool.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Voice-over&lt;/strong&gt; with Cartesia or ElevenLabs, or with your own audio files (from any TTS, local models included). The video waits for each sentence, so the voice sets the pace.&lt;/li&gt;
&lt;/ul&gt;

&lt;h2&gt;
  
  
  Try it
&lt;/h2&gt;



&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight shell"&gt;&lt;code&gt;npx agent-screenplay init
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;&lt;code&gt;init&lt;/code&gt; downloads Chromium once, adds the MCP server to Claude Code, Codex and Cursor, and installs the guide that teaches the agent to write good scenarios. Restart your agent, then ask for a video.&lt;/p&gt;

&lt;p&gt;It's brand new, so I'd love feedback, especially on the scenario format and the default look. Issues and stars are very welcome: &lt;a href="https://github.com/gmpetrov/agent-screenplay" rel="noopener noreferrer"&gt;https://github.com/gmpetrov/agent-screenplay&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;&lt;em&gt;Written with an AI agent and published at my request; every fact comes from the repo.&lt;/em&gt;&lt;/p&gt;

</description>
      <category>showdev</category>
      <category>opensource</category>
      <category>ai</category>
      <category>typescript</category>
    </item>
    <item>
      <title>What do you think of this SaaS template ?</title>
      <dc:creator>Georges Petrov</dc:creator>
      <pubDate>Sun, 26 Sep 2021 13:05:14 +0000</pubDate>
      <link>https://dev.to/gmpetrov/what-do-you-think-of-my-template-60p</link>
      <guid>https://dev.to/gmpetrov/what-do-you-think-of-my-template-60p</guid>
      <description>&lt;p&gt;Got tired of having to implement auth and payment on my side-projects.&lt;/p&gt;

&lt;p&gt;I'made this SaaS business template to focus on what really brings value to my customers.&lt;/p&gt;

&lt;p&gt;Made with Typescript/React.js/Next.js/NextAuth.js/Prisma/Stripe/Tailwindcss/Postgresql&lt;/p&gt;

&lt;p&gt;&lt;a href="https://github.com/gmpetrov/ultimate-saas-ts" rel="noopener noreferrer"&gt;https://github.com/gmpetrov/ultimate-saas-ts&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;What features would you add to make it more useful ?&lt;/p&gt;

</description>
      <category>react</category>
      <category>typescript</category>
      <category>nextjs</category>
      <category>tailwindcss</category>
    </item>
  </channel>
</rss>
