# I built Web-MCP-Bridge: An open-source extension and Python bridge to run local MCP tools in web AI interfaces > Source: > Published: 2026-11-03 11:12:25+00:00 Hey everyone, I built Web-MCP-Bridge, a 100% open-source browser extension and local Python WebSocket bridge designed to bring Model Context Protocol (MCP) tool execution directly to web-based AI chat interfaces (such as ChatGPT, Gemini, Claude, DeepSeek and other browser chats). The Problem: Most power users want to give web AI interfaces access to local tools (filesystems, local terminal execution, custom automation scripts, or desktop management). However, web chat UIs don't natively support local MCP execution, often requiring users to set up paid API tokens or run heavy standalone apps. How it works: Tool Interception: The extension intercepts structured JSON tool calls inside the browser chat UI. Local Bridge: It routes the request over a local WebSocket (127.0.0.1) to a lightweight Python process. Execution and Return: The bridge executes the command on your local stdio MCP server (like astrax-mcp or any custom local server) and returns the output back into the chat window. Key Highlights: Zero API Requirements: Works directly with your standard browser chat sessions. Granular Security: 3-state permission model (Allow always, Ask every time, Don't allow) requiring user confirmation before executing any local tool. Local-First and Private: All WebSocket communication stays strictly on 127.0.0.1. Fully Open-Source: Both the browser extension (Manifest V3) and Python bridge server are open-source. Stack: Extension: JavaScript (Manifest V3, Content Scripts, Service Workers) Local Bridge: Python (WebSockets, asyncio, stdio MCP execution) Check out the code, run it locally, and contributions or feedback are very welcome! GitHub Repository:[https://github.com/AstraxOf/web-mcp-bridge](https://github.com/AstraxOf/web-mcp-bridge)