M
MCP Directory CURATED DIRECTORY
911 MCP servers indexed
中文
← Back to directory
B

Browser Use

Official
Make websites accessible for AI agents. Automate tasks online with ease.
How to install ↓ Compare GitHub source repository ↗
Category
Browser #31 of 32
Stars
★ 117.7k Very popular
Transport
stdio (local process)
Runtime
Python 3.11+ · Prebuilt binary
Credentials
Optional API key
License
MIT
Last commit

Static review · not runListed 2026-08-07

FMRS 33/100 · D · Weakest Reliability 5/20 — Read the FMRS scoring method →

Fit and risk

What it can accessUses the networkControls a browser

Best for

  • Users who want AI agents to perform one-off browser tasks (e.g., via Claude Code, Cursor, etc.)
  • Developers building software that automates the web using the Python library for repeatable, scalable tasks

Not for

  • Scenarios not requiring browser automation or simple API calls
  • Simple tasks where cost sensitivity is high and background browser services are unnecessary

Required permissions

  • Access to local Chrome browser process
  • Read and manipulate web page content (forms, buttons, data)
  • Potential read access to local files (e.g., configuration)
  • Network access to load web pages and call LLM APIs

Risks and side effects

  • May perform unintended web actions (e.g., submitting forms, purchasing items)
  • Requires LLM API keys, which may incur costs
  • Automated browsing can be detected by websites and lead to IP bans
  • Handling sensitive data (like login credentials) requires privacy consideration

Setup

Before you start

Runtime：Python 3.11+ · Prebuilt binary

BROWSER_USE_API_KEY optionalsecret API key for Browser Use Cloud, grants access to all supported models; get it at cloud.browser-use.com/new-api-key.
GOOGLE_API_KEY optionalsecret Optional bring-your-own key for using Google Gemini models.
ANTHROPIC_API_KEY optionalsecret Optional bring-your-own key for using Anthropic Claude models.

Install with pip: pip install browser-use, then launch the MCP server via uvx: uvx browser-use --cli-mcp. Alternatively, follow the quickstart prompt in the README to let your agent auto-install and register the skill.

Check that it works

Launch via the manifest command `uvx browser-use --cli-mcp` and confirm the server connects in your MCP client; send a simple browser task (e.g. 'open browser-use.com and tell me the page title') — if the agent controls a real Chrome browser and completes it, the connection works.

Troubleshooting

  1. Ensure Python version >= 3.11
  2. Check that LLM API keys are installed and configured correctly (e.g., BROWSER_USE_API_KEY)
  3. Verify Chrome browser is installed and executable
  4. Inspect browser console or logs for error messages
  5. For CAPTCHA or anti-bot issues, consider using Browser Use Cloud

Things to try

Once connected, you can ask your AI assistant things like:

  • Open Chrome and extract structured data about my followers, then export it as a CSV
  • Fill in this job application with my resume and information
  • Compare these three laptops and give me a table with prices
  • QA test my local website and report any bugs and usability issues

Use cases

Fill in a job application with resume and information
Extract structured data about followers and export as CSV
QA test a local website and report bugs, usability issues, and visual inconsistencies

Overview

Browser Use is an open-source Python library and MCP server that enables AI agents to control a real Chrome browser just like a human: open pages, click buttons, type, fill forms, and extract data. You describe the task, and the agent completes it. It supports various LLM providers and optionally integrates with Browser Use Cloud for large-scale, stealthy execution.

Similar servers

Compare with other MCP servers

Source revision c75e8476e26d Data synced 2026-10-12 Read the FMRS scoring method