browser-use-mcp-server
An MCP server that lets an agent drive a real browser: open pages, click and fill forms through Playwright
- 848 ★ on the repository
- Last commit
- MIT
Extended access
Why this level
- Drives a real browser and makes network requests
- Acts on sites on your behalf under model control
Install
- Works in
Manual install
uv tool install mcp-proxy
uv pip install playwright
uv run playwright install --with-deps --no-shell chromium
uv build
uv tool install dist/browser_use_mcp_server-*.whlBuild and install the server via uv, then run browser-use-mcp-server.
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- Drives a real browser and makes network requests
- Acts on sites on your behalf under model control
This is third-party code. Review the repository files before installing.
What it does
The server wraps a browser agent built on the browser-use library and Playwright and exposes it over MCP. Through it an assistant opens sites, follows links, clicks and fills fields on your behalf. It runs in two transports, SSE and stdio, with ready configs for Cursor, Windsurf and Claude. A language model drives the browser actions, so an OpenAI key is required.
Who it is for. For developers and testers who want an agent to operate a browser on its own.
Good fit when
- You want the agent to navigate a site and fill forms
- You need a browser inside Cursor, Windsurf or Claude as an MCP tool
- You want an isolated browser environment via Docker
Not a fit when
- You have no OpenAI key and cannot use an external model
- Browser actions must stay strictly manual, without an autonomous agent
Example request
Open the site, find its pricing page and list the plansMCP
- Transport
- sse, stdio
- Authentication
- API key
Environment variables
OPENAI_API_KEYrequired, secret- OpenAI key: the model drives the browser actions.
Limitations
It needs an OpenAI key, and that API is not reachable from Russia without a VPN. It requires Playwright and a Chromium browser, and Docker helps isolate it. The agent acts in the browser on its own, so run it only on your own accounts and resources and watch what it does.
How to disable. Remove the server entry from your MCP client config and uninstall the tool with uv tool uninstall browser-use-mcp-server.
FAQ
Which transports are supported?
SSE and stdio. For SSE the client connects by URL; for stdio the server runs as a command with arguments.
Why an OpenAI key?
A language model drives the browser actions, deciding where to click and what to type, so an external key is needed.
Discussion
No questions or reviews yet.
Related
MCP serversAn MCP wrapper over the browser-use engine: a single tool drives a browser from a natural-language task.
An MCP server and Chrome extension that let the agent drive your already open browser with your logged-in sessions, all locally
MCP Server Playwright by Automata Labs
MCP Server Playwright
Playwright-based MCP server for browsing sites, clicking, filling forms, taking screenshots and running JavaScript in a browser
Microsoft's official MCP server for driving a browser with Playwright through the accessibility tree, no screenshots needed
MCP gives an agent a Playwright browser where a person can take over the same session and save an authentication profile
A Chrome extension with an MCP server: the agent drives your everyday browser with its sessions, tabs, bookmarks and history
Comparison with similar tools
Cards that share the main search query, and how they differ.
| Tool | Official | Authentication | Repository stars | Main difference |
|---|---|---|---|---|
| browser-use-mcp-server (this card) | no | API key | 848 | |
| MCP server for 1C:Naparnik | no | API key | 47 | on Spring Boot, for the 1C:Naparnik API |
| Kommo CRM MCP server | no | OAuth | 0 | in Python, with OAuth for Kommo |
| MCP server for the GigaChat API | no | API key | 0 | a small server for Sber's GigaChat API |
README in short
The README covers installation via uv, Playwright with Chromium, and running in SSE or stdio mode. It lists client configs for Cursor, Windsurf and Claude with the paths to their settings files. It offers a Docker run with mapped ports and an OPENAI_API_KEY variable. The project is Python, MIT licensed.