ChromeDevTools/chrome-devtools-mcp

Chrome DevTools for agents (chrome-devtools-mcp)

What it is – A Node‑based server that exposes Chrome DevTools functionality over the Model‑Context‑Protocol (MCP). It lets an AI coding assistant (Claude, Copilot, Cursor, Antigravity, etc.) drive a real Chrome browser: inspect pages, record performance traces, debug network traffic, take screenshots, and automate actions with Puppeteer.

Why it matters – Modern AI agents are good at generating code but still need a reliable way to interact with a live browser. By speaking MCP, the agent can ask the tool to “open this URL and give me a performance report” or “click this button and return the console output” without writing custom Selenium scripts. The server handles the heavy lifting (launching Chrome, waiting for results, translating DevTools data into JSON) so the agent can focus on higher‑level reasoning.

Key capabilities

• Performance insights – Records Chrome traces, extracts metrics, and (optionally) augments them with real‑user data from the CrUX API.
• Debugging utilities – Access network logs, console messages (with source‑mapped stack traces), and screenshots.
• Reliable automation – Wraps Puppeteer to perform clicks, navigation, form filling, etc., with automatic waiting for page stability.
• MCP interface – Exposes all of the above via a standard JSON‑over‑WebSocket protocol that AI agents can call as tools.
• CLI mode – A standalone command‑line tool (npx chrome-devtools-mcp) for ad‑hoc use without an MCP client.

How to get started

1. Install Node LTS and a recent stable Chrome.
2. Add the server to your agent’s MCP configuration, e.g.

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

3. In the agent’s prompt, ask a task such as Check the performance of https://developers.chrome.com. The server will launch Chrome, record a trace, and return the analysis.
4. For lightweight use, start the server with --slim --headless.

Configuration & advanced use – Flags control headless mode, isolated user data directories, concurrent sessions, connecting to an existing Chrome instance, or Android debugging. See the linked docs for the full list.

Privacy & data – By default the tool sends anonymous usage stats and may query the CrUX API for field performance data. Both can be disabled with --no-usage-statistics or the environment variables CHROME_DEVTOOLS_MCP_NO_USAGE_STATISTICS / CI.

Typical users – Developers building AI‑powered coding assistants, IDE extensions, or autonomous agents that need to browse the web, profile page speed, or debug front‑end code.

---

All details are taken directly from the repository’s README.