chromectx

For those who routinely end up with 90+ Chrome tabs. Search all your open tabs by title or full page content, save entire windows to JSON, restore them later, and expose everything as RAG context for AI tools via MCP.

# Fast — searches tab titles and URLs only
chromectx search 'figma design system'

Instantly finds the right tab across 90+ open tabs. No page fetching — results appear in milliseconds.

# Deep — fetches and indexes full page content
chromectx search 'max connections pool timeout' --deep

You read something 20 tabs ago about connection pool limits causing timeouts in production — but which tab was it? The title just says "Configuration" or "Advanced Settings." --deep fetches every open page's content and searches the actual text, so you find it without retracing your steps.

And even better: make your open browser tabs RAG context for Claude, Cursor, and other AI tools via the Model Context Protocol.

Install

npx chromectx install

This one command:

1. Checks your environment (macOS, Chrome, automation permissions)
2. Installs chromectx globally
3. Offers to configure the MCP server for your AI tools (Claude Desktop, Cursor, VS Code, etc.)

Manual setup

If you prefer to set things up yourself:

npm i -g chromectx          # install the CLI globally
chromectx setup             # configure MCP for your AI tools

Or add the MCP server config directly:

{
  "mcpServers": {
    "chromectx": {
      "command": "npx",
      "args": ["-y", "chromectx@latest", "mcp"]
    }
  }
}

Requirements

• macOS (uses AppleScript/JXA to communicate with Chrome)
• Google Chrome
• Node.js >= 18
• Automation permission: your terminal must be allowed to control Chrome (System Settings > Privacy & Security > Automation)

Commands

| Command | Description | | -------------------------- | ------------------------------------------------------------- | | chromectx install | One-time guided setup (env check, global install, MCP config) | | chromectx setup | Configure MCP server for AI tools | | chromectx doctor | Check runtime environment and report readiness | | chromectx list | List open Chrome windows and tabs, or saved sessions | | chromectx search <query> | Search tab titles and URLs; --deep searches page content | | chromectx save | Save a Chrome window as a session file | | chromectx restore | Restore a saved session to Chrome | | chromectx rag <query> | RAG search across open tabs (JSON output for agents/scripts) | | chromectx mcp | Start the MCP server over stdin/stdout |

Global flags

-h, --help     Show help
--version      Print version
--json         Structured JSON output (where supported)
-q, --quiet    Reduce output
-v, --verbose  Increase diagnostic output

MCP tools

When running as an MCP server, chromectx exposes:

• **rag_chrome_search** -- Search the content of all open Chrome tabs using BM25 ranking. Fetches and indexes every open tab, then runs a keyword search. Supports chunk-level or full-page results.
• **doctor** -- Inspect the local macOS runtime and report whether chromectx is ready to run.

Updating

The MCP server auto-updates: each restart fetches the latest version via npx.

For the global CLI, re-run:

npx chromectx install

chromectx will also notify you in the terminal when a new version is available.

License

MIT License

Copyright 2025-2026 Anubhav Saha