Webmcp Bridge

WebMCP Bridge lets web applications expose their functionality as tools that AI assistants can discover and call. It reads tools registered on pages via the navigator.modelContext browser API and makes them available to desktop AI clients like Claude Code and Cursor through the Model Context Protocol (MCP).

How it works:

- Your web app registers tools using a library like webmcp-react
- The extension detects those tools on activated pages
- Tools from all active tabs are aggregated and exposed through a local MCP server
- AI assistants connect to the server and can call your app's tools directly

Features:

- Per-tab activation ("Until reload") or per-domain activation ("Always on")
- Tools are namespaced by tab to avoid collisions across pages
- Status indicator shows connection state and active tool count
- Supports any page that implements the navigator.modelContext API

Setup:

1. Install this extension
2. Add the MCP server to your AI client (see instructions: https://github.com/MCPCat/webmcp-react/blob/main/extension/README.md)
3. Open a page with registered tools and activate the extension
4. Your AI assistant can now see and call the page's tools

The companion MCP server (webmcp-server) runs locally as a stdio process. Install it with: npx webmcp-server

All communication stays on your machine. The extension connects to the local server over ws://127.0.0.1 and never sends data externally.

Open source under the MIT license: https://github.com/mcpcat/webmcp-react