Browser-Use MCP
Open-source Model Context Protocol server exposing Browser-Use vision navigation to Claude Desktop and local LLMs.
Claude Desktop Configuration
Add this snippet to your claude_desktop_config.json to enable full browser capabilities:
{
"mcpServers": {
"browser-use": {
"command": "uvx",
"args": [
"browser-use-mcp"
],
"env": {
"OPENAI_API_KEY": "YOUR_OPENAI_KEY"
}
}
}
}uvx browser-use-mcpExposed Tools & Primitives (5)
These functions are invoked autonomously by Claude, Cursor, or your AI agent loop:
browser_stepbrowser_task_runbrowser_get_interactive_elementsbrowser_click_coordinatebrowser_extract_contentAI Compatibility & Anti-Bot Specifications
| Claude Desktop Integration | ✔ Yes — Fully Tested with Claude 3.5 Sonnet / Haiku |
|---|---|
| Cursor IDE / Windsurf | ✔ Yes — Standard Stdio / SSE Transport |
| Custom Agent Frameworks | ✔ Supported (LangChain, LlamaIndex, CrewAI) |
| Automated CAPTCHA Solving | LLM vision reasoning; external solver plugin |
| Stealth & Evasion Engine | Playwright stealth integration |
| Pricing & Licensing | 100% Free and open source (BYO LLM keys) |
| Source Code Repository | https://github.com/browser-use/browser-use ↗ |
Strengths for AI Agents
- ✔ Vision-guided element labeling enables robust interaction
- ✔ Huge open-source community support and fast iterations
Considerations
- ✖ Consumes significant LLM tokens for vision analysis per step
Evaluation & Benchmark Verdict
Browser-Use MCP allows LLM clients to control web browsers using vision-guided agent execution.
Our Verdict
Browser-Use MCP brings vision-based multi-step agent navigation to any MCP-compliant client.
Ideal Use Case
Complex multi-step web workflows where traditional DOM selectors break frequently.
Avoid If
Low-cost high-frequency data extraction.
Powered by Browser-Use
Open-source and managed cloud browser infrastructure created to make websites accessible for LLM agents.
Alternative Browser MCP Servers
2Captcha Browser MCP
The premier Model Context Protocol (MCP) server for autonomous AI agents with native CAPTCHA solving, stealth browsing, and proxy rotation.
Browserbase MCP Server
Official Model Context Protocol server connecting Claude and AI agents to Browserbase cloud sessions with live inspection.
Kernel MCP Server
High-speed Model Context Protocol server executing ephemeral cloud browser microVMs for Claude and Cursor.