Browser Tools Mcp

BrowserTools MCP, from AgentDesk, gives an AI coding agent visibility into the developer's own Chrome session. It streams console output, network activity, screenshots and Lighthouse audits from the browser tab the developer already has open, and already logged into, to any MCP compatible client, including Cursor, Claude Code, Windsurf, Cline, Zed and Gemini CLI. It is built for developers debugging a live application rather than writing automated tests. Browser automation servers start a fresh browser, and recent Chrome versions block remote debugging on the default profile, so login state has to be rebuilt before debugging. BrowserTools avoids that by attaching through a DevTools extension. Setup is one npx command in the MCP client configuration plus the unpacked Chrome extension, and it needs Node 22.19 or newer. Tools cover console logs and errors, network logs and failed requests, the selected element, page info, tab listing, screenshots returned as images, browser storage, and Lighthouse accessibility, performance, SEO and best practice audits. Built in prompts supply debugging and audit workflows, large payloads such as HAR files and full audit reports are exposed as MCP resources rather than inlined, and every read tool supports paging and keyword filters to keep responses small. BrowserTools MCP is free and open source, published on npm and GitHub. No pricing is published for it. Version 2.0 is a rewrite, and the project urges anyone on 1.2.x to upgrade because that release has a critical vulnerability. The new version binds only to loopback, requires a per run token, and scrubs credentials such as authorization headers, cookies and tokens before data leaves the browser. AgentDesk also lists an AgentDesk Studio product as coming soon, with a waitlist.