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.
To guarantee complete privacy and zero server overhead, the entire extension runs serverless:
Framework: Plasmo (React + TypeScript + Chrome MV3). Storage: chrome.storage.local handles the entire project matrix. No external databases.
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.
Seamless Context Injection
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.
Check out the source listing or download it free on the Chrome Web Store if you want to optimize your multi-AI dev loops!
🔗 Chrome Web Store: https://chromewebstore.google.com/detail/jiepmjajdlemomagfihinmkbfaednncf?utm_source=item-share-cb