Browser
by BrowserMCP
Browser MCP puts an agent in the driver's seat of your actual, already-logged-in Chrome window via a companion extension, instead of spinning up a fresh browser instance.
Add to Claude Code
claude mcp add browsermcp -- npx @browsermcp/mcp@latest
About
Browser MCP pairs an MCP server with a Chrome extension so tools like Claude, Cursor, VS Code, and Windsurf can automate the browser you're already using, not a freshly launched, logged-out instance. Because it drives your real browser, it inherits your existing sessions and cookies, so an agent can act on sites you're already signed into without a separate login flow.
It's adapted from Microsoft's Playwright MCP server, keeping a similar tool set for navigation, clicking, typing, and accessibility snapshots, but pointing those tools at your live browser profile rather than a sandboxed one, which also means it inherits your real fingerprint and is less likely to trip basic bot detection or CAPTCHAs than a freshly spawned automated browser.
Key features
- Automates your existing, already-logged-in Chrome window instead of a fresh instance
- Runs entirely locally, no browser activity leaves your machine
- Inherits your real browser fingerprint, reducing exposure to basic bot detection
- Playwright-MCP-derived tool set: navigate, click, drag, take screenshots, capture accessibility snapshots
- Works with Claude, Cursor, VS Code, and Windsurf out of the box
Use cases
- Automating a workflow on a site you're already logged into without re-authenticating a separate session
- Filling out and submitting forms using your real browser profile
- Extracting structured content from pages via accessibility snapshots rather than screenshots
- Running lightweight LLM-driven testing or QA against your actual dev environment in the browser
Available tools
navigate
Navigates the browser to a specified URL.
click
Clicks an element on the current page, identified via the accessibility snapshot.
snapshot
Captures an aria-level accessibility snapshot of the current page, including structure and metadata.
screenshot
Takes a screenshot of the current browser view.
drag
Drags an element and drops it onto another.
getConsoleLogs
Retrieves the browser's console log output.
Frequently asked questions
Why automate my real browser instead of a fresh headless one?
So the agent inherits your existing logins and sessions automatically, and uses your actual browser fingerprint, making it less likely to get flagged by basic bot detection than a freshly spawned automation profile.
Is my browsing data sent anywhere?
No. Automation and data stay local to your machine; the extension and MCP server communicate directly without a remote relay.