MCP server w/ Browser Use
@JovaniPink
FastAPI server implementing MCP protocol Browser automation via browser-use library.
Overview
What is MCP server w/ Browser Use?
MCP server w/ Browser Use is a Model Context Protocol server that wires the browser-use automation engine into Claude Desktop and other MCP-compatible clients. It exposes a single run_browser_agent tool that orchestrates a browser session, executes browser-use agent, and returns the result.
How to use MCP server w/ Browser Use?
Install by cloning the repository and running uv sync. Copy .env.example to .env and set your API keys and configuration values. Launch the server with uv run mcp-browser-use. For Claude Desktop, add the server to your MCP configuration with environment variables for the LLM provider and model.
Key features of MCP server w/ Browser Use
- Automated browsing via natural-language instructions.
- Agent lifecycle management with history export and error handling.
- Centralized browser configuration using environment variables.
- FastMCP integration for session cleanup and tool registration.
- Client helpers for in-process testing with async Python.
Use cases of MCP server w/ Browser Use
- Automate web interactions through MCP clients like Claude Desktop.
- Perform form filling, navigation, and data extraction on websites.
- Run headless browser tasks for testing or scraping.
- Integrate browser automation into AI agent workflows.
FAQ from MCP server w/ Browser Use
What dependencies are required?
Python 3.11+, Google Chrome or Chromium, and uv for dependency management.
How is the LLM configured?
Set environment variables MCP_MODEL_PROVIDER, MCP_MODEL_NAME, MCP_TEMPERATURE, MCP_MAX_STEPS, MCP_MAX_ACTIONS_PER_STEP, and MCP_USE_VISION. Provider-specific API keys (e.g., ANTHROPIC_API_KEY, OPENAI_API_KEY) are also required.
Where does browser data persist?
The browser session can use persistent profiles via CHROME_PERSISTENT_SESSION and supports proxies and custom Chromium flags.
Can I run tests?
Yes, use uv run pytest to run the unit tests covering agent behavior, browser session factory, and utility helpers.
What security considerations exist?
Controlling a full browser instance remotely can grant broad access to the host machine. Review the SECURITY.md documentation before exposing the server to untrusted environments.