A Docker container that bundles Claude Code, a web UI, a headless Chromium browser, several AI coding CLIs and 50+ dev tools in one ready-to-run workstation.
AgentDeskAI/browser-tools-mcp
An MCP server and Chrome extension that give AI coding agents access to console logs, network traffic, screenshots and Lighthouse audits from your logged-in Chrome session.
- Stars
- 7,332
- Forks
- 536
- Open issues
- 2
- Last commit
- 9 days ago
What is it?
BrowserTools MCP lets an AI coding agent inspect the page you are actually working on. It attaches to your existing Chrome session through a DevTools extension, so you stay logged in, and streams console output, network activity, screenshots and Lighthouse audit results to any MCP-compatible client such as Cursor, Claude Code, Windsurf, Cline, Zed or Gemini CLI.
The project targets debugging of live apps rather than writing automated tests. Automation-focused servers such as Chrome DevTools MCP or Playwright MCP drive a fresh browser profile, which means recreating login state before debugging. BrowserTools instead reads the session you already have open. Setup is one MCP server command plus loading the bundled extension as an unpacked Chrome extension; the server runs the connector itself.
The server binds to loopback only, requires a per-run token for the HTTP API, and scrubs credentials such as Authorization and Cookie headers and common token formats from captured data. Multiple tabs with DevTools open are tracked separately, and large payloads are exposed as MCP resources so they only enter the agent's context when requested.
Key features
- 01Streams console logs, errors and network requests from the current Chrome tab.
- 02Captures screenshots and returns them as images to the agent.
- 03Runs Lighthouse accessibility, performance, SEO and best-practices audits.
- 04Scrubs credentials from captured headers and strings before they leave the browser.
- 05Tracks multiple tabs separately and lets tools target a specific tab ID.
- 06Supports filtering and paging of log results to keep responses small.
Use cases
- Asking an agent to diagnose console errors on a page you are logged into.
- Debugging failed or slow API requests while developing a web app locally.
- Running Lighthouse audits on a page without setting up a separate test profile.
- Having an agent check a UI change visually by reading a screenshot of the live page.
Quick start
- 1
Add the MCP server to your MCP client config
npx -y @agentdeskai/browser-tools-mcp@latest - 2
Download or clone the repository
- 3
In chrome://extensions, enable Developer mode and load the chrome-extension directory unpacked
- 4
Open Chrome DevTools (F12) on the page you want to inspect
Things to know
- Requires Node 22.19 or newer.
- Firefox is not verified and is treated as unsupported.
- Network capture only records requests made after DevTools is opened.
In collections
Topics
FAQ
- What is browser-tools-mcp?
- An MCP server and Chrome extension that give AI coding agents access to console logs, network traffic, screenshots and Lighthouse audits from your logged-in Chrome session. BrowserTools MCP lets an AI coding agent inspect the page you are actually working on. It attaches to your existing Chrome session through a DevTools extension, so you stay logged in, and streams console output, network activity, screenshots and Lighthouse audit results to any MCP-compatible client such as Cursor, Claude Code, Windsurf, Cline, Zed or Gemini CLI.
- Is browser-tools-mcp free to use? What license does it have?
- browser-tools-mcp is open source under the MIT license. Check the license terms before using it in commercial products.
- Is browser-tools-mcp still maintained?
- Actively maintained. The last commit to browser-tools-mcp was on Oct 16, 2026.
- What can I use browser-tools-mcp for?
- Asking an agent to diagnose console errors on a page you are logged into. Debugging failed or slow API requests while developing a web app locally. Running Lighthouse audits on a page without setting up a separate test profile. Having an agent check a UI change visually by reading a screenshot of the live page.
- How do I install browser-tools-mcp?
- Add the MCP server to your MCP client config: npx -y @agentdeskai/browser-tools-mcp@latest Download or clone the repository In chrome://extensions, enable Developer mode and load the chrome-extension directory unpacked Open Chrome DevTools (F12) on the page you want to inspect
- Where can I get browser-tools-mcp?
- The source code and installation instructions are on GitHub: https://github.com/AgentDeskAI/browser-tools-mcp
Summary generated by AI from the project's README. GitHub stats refresh daily.
Related projects
CLI proxy that filters and compresses shell command output before it reaches an AI coding agent's context, cutting token use.
A Claude Code plugin that adds a live status line showing context usage, rate limits, active tools, running agents, and todo progress.
Open-source Rust engine that filters and compresses the context AI coding tools read, with MCP server, shell hooks and local audit records.
A CLI and web tool that packs an entire code repository into one AI-friendly file for feeding to LLMs like Claude or ChatGPT.
A lightweight cross-platform IDE for AI coding that runs multiple Claude Code and Codex agent sessions across projects in one window.