<?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: ContextOS</title>
    <description>The latest articles on DEV Community by ContextOS (@contextosai).</description>
    <link>https://dev.to/contextosai</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.amazonaws.com%2Fuploads%2Fuser%2Fprofile_image%2F3982134%2F5a2792fb-423a-41ae-9905-818e633d70ca.png</url>
      <title>DEV Community: ContextOS</title>
      <link>https://dev.to/contextosai</link>
    </image>
    <atom:link rel="self" type="application/rss+xml" href="https://dev.to/feed/contextosai"/>
    <language>en</language>
    <item>
      <title>How I Built a Local-First Chrome Extension to Chain ChatGPT, Grok, and Claude</title>
      <dc:creator>ContextOS</dc:creator>
      <pubDate>Sat, 13 Jun 2026 04:15:27 +0000</pubDate>
      <link>https://dev.to/contextosai/how-i-built-a-local-first-chrome-extension-to-chain-chatgpt-grok-and-claude-3km2</link>
      <guid>https://dev.to/contextosai/how-i-built-a-local-first-chrome-extension-to-chain-chatgpt-grok-and-claude-3km2</guid>
      <description>&lt;p&gt;Shifting between LLMs for their unique strengths is standard workflow now, but managing the data state between browser tabs is an absolute nightmare. Here is how I used the Plasmo framework to build a lightweight, local-first context compiler called ContextOS.&lt;/p&gt;

&lt;p&gt;To guarantee complete privacy and zero server overhead, the entire extension runs serverless:&lt;/p&gt;

&lt;p&gt;Framework: Plasmo (React + TypeScript + Chrome MV3).&lt;/p&gt;

&lt;p&gt;Storage: chrome.storage.local handles the entire project matrix. No external databases.&lt;/p&gt;

&lt;p&gt;The Parsing Loop: Users input a free Gemini API key (BYOK). When a user hits ✨ Sync to Project Context, a background worker handles the raw text payload, processes it structurally, and commits it into dynamically named local state blocks.&lt;/p&gt;

&lt;p&gt;Seamless Context Injection&lt;br&gt;
The hardest part was getting clean injection mechanics across wildly different DOM layouts (ChatGPT, Grok, Claude, etc.). I implemented weightless content scripts that target the active webpage textareas, formatting the project's modular buckets into clean, hierarchical Markdown right before injection.&lt;/p&gt;

&lt;p&gt;Check out the source listing or download it free on the Chrome Web Store if you want to optimize your multi-AI dev loops!&lt;/p&gt;

&lt;p&gt;🔗 Chrome Web Store: &lt;a href="https://chromewebstore.google.com/detail/jiepmjajdlemomagfihinmkbfaednncf?utm_source=item-share-cb" rel="noopener noreferrer"&gt;https://chromewebstore.google.com/detail/jiepmjajdlemomagfihinmkbfaednncf?utm_source=item-share-cb&lt;/a&gt;&lt;/p&gt;

</description>
      <category>ai</category>
      <category>webdev</category>
      <category>productivity</category>
      <category>web</category>
    </item>
  </channel>
</rss>
