Allows an AI agent to access and interact with Gmail through a locally controlled Chrome browser, using the user's logged-in session without modifying the real Chrome profile.
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type @ followed by the MCP server name and your instructions, e.g., "@browser-mcp-cdp check my Gmail inbox"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
A Model Context Protocol (MCP) server that drives your real local Chrome via the Chrome DevTools Protocol. Designed for agents that need access to your logged-in sessions (Gmail, internal dashboards, banking, etc.) without uploading credentials to a cloud service.
Different tradeoffs from the alternatives:
browser-mcp-cdp | Playwright MCP | Chrome DevTools MCP | |
Browser | Your local Chrome | Managed Playwright Chromium | Your local Chrome |
Login state | ✅ Snapshot of your profile (cookies, logins, IndexedDB) | ❌ Fresh profile | ✅ Real profile (modifies it) |
Profile isolation | ✅ Copy-on-start, real profile untouched | N/A | ⚠️ Direct attach |
Multi-client | ✅ Shared broker across MCP clients | ❌ Per-client | ❌ Per-client |
Dependencies | Node only (no Playwright download) | Playwright + browsers | Node only |
The profile snapshot pattern means you can ask an agent to "check my email" — it sees your Gmail logged in — but the agent's actions never modify your real Chrome profile. Each session gets a disposable copy.
Related MCP server: CloakBrowser MCP
claude mcp add browser -- npx -y browser-mcp-cdpOr add manually to ~/.claude.json / Claude Desktop config:
{
"mcpServers": {
"browser": {
"command": "npx",
"args": ["-y", "browser-mcp-cdp"]
}
}
}Requires Node 18+ and Google Chrome installed.
Tool | Purpose |
| Load a URL, return final URL + title |
| Run JS in the page, return result |
| PNG of current page |
| Click via CSS selector |
| Click at pixel coordinates (pierces iframes / overlays) |
| Set value on input/textarea/select |
| Wait for a selector to appear |
| Current URL |
| Visible text of page or element |
|
|
| Scroll by |
| List open tabs |
| Open a new tab |
Var | Default |
| Auto-detected per OS |
| OS-standard Chrome profile dir |
|
|
|
|
|
|
First MCP client connection spawns a broker subprocess (detached).
Broker copies your Chrome profile's cookies/local storage/login data to a temp dir.
Broker launches Chrome with --user-data-dir=<temp> and --remote-debugging-port.
Subsequent MCP clients connect to the same broker over a Unix socket → they share one Chrome.
Broker exits after BROKER_IDLE_TIMEOUT_MS of no activity; temp profile is deleted.
This server lets an LLM execute arbitrary JavaScript in a browser that is logged into your accounts. Only use with models and prompts you trust. Profile snapshotting limits damage (no persistent mutations to your real Chrome), but the agent can read cookies, session tokens, and any data visible to logged-in you for the duration of the session.
MIT
Click an element matched by a CSS selector. Returns true if found and clicked.
Inputs: selector (required).
Click at an absolute (x, y) pixel coordinate on the page using real CDP mouse events. Pierces through iframes and ad overlays that block CSS-selector clicks.
Inputs: x (required), y (required).
Execute JavaScript in the current page context and return the result. Use await_promise=true (default) for async expressions.
Inputs: expression (required), await_promise.
Set the value of an input, textarea, or select element matched by a CSS selector.
Inputs: text (required), selector (required).
Get the outer HTML of the page or a specific element.
Inputs: selector.
Get the visible text content of the page or a specific element.
Inputs: selector.
Get the current page URL.
Inputs: no named parameters.
Open a new browser tab, optionally navigating to a URL.
Inputs: url.
Capture a screenshot of the current page. Returns a PNG image.
Inputs: no named parameters.
Scroll the page by a relative offset in pixels.
Inputs: x, y.
List all open browser tabs with their id, url, and title.
Inputs: no named parameters.
Wait until an element matching the CSS selector appears in the DOM. Returns true if found within the timeout, false otherwise.
Inputs: selector (required), timeout_ms.
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
Scored across 13 tools
Each tool targets a distinct action (navigate, click, fill, etc.) with clear boundaries. No two tools do the same thing; even click variants differ by selector vs coordinates.
All tools use the 'browser_' prefix with a consistent verb_noun pattern (e.g., get_url, click_at). The only outlier is 'browser_tabs' which uses a noun instead of a verb, but it's still clear.
13 tools is well-scoped for browser automation, covering navigation, interaction, content extraction, scrolling, and tab management without being overwhelming.
Covers most core browser operations (navigate, click, fill, read, screenshot, JS eval). Minor gaps like back/reload or cookie handling exist but are not critical for typical use cases.
Give AI agents a real cloud Chrome: open sessions, navigate, read pages, click, type and screenshot.
Let your AI agent operate any website through a real, logged-in browser session.
Let AI agents use your real, signed-in Chrome: tabs, clicks, forms, screenshots and video.
Control a real browser from AI agents: navigate, click, type, screenshot, read the page.