Automation
browser-use avatar

browser-use

AI-powered browser automation server for web interaction, data extraction, and research using the Model Context Protocol.

Introduction

This MCP server provides a powerful interface for AI assistants to control a web browser, enabling complex web automation and data extraction workflows. By wrapping the browser-use library with an HTTP-based transport layer, it solves common timeout issues associated with standard stdio communication, allowing for long-running browser tasks that last 30 to 120 seconds or more. It is designed for developers and power users who need to automate interactions, perform multi-source research, or build reusable browser workflows.

  • Full browser automation: Navigate pages, click elements, fill forms, and interact with complex web UI components programmatically.

  • Deep research capabilities: Automatically conduct multi-source web searches, synthesize findings, and generate structured markdown reports.

  • Skill management system: Learn, save, and manage custom browser-based skills, allowing for repeatable and parameterizable workflows across sessions.

  • Persistent task monitoring: Use the built-in HTTP daemon to run tasks in the background, view real-time progress, and retrieve results via a web-based dashboard.

  • Observability and logging: Gain full visibility into execution logs, task success rates, and system health status for improved reliability and debugging.

  • The agent relies on an LLM (such as Gemini, GPT-4o, or Claude) to interpret high-level natural language instructions and decide on browser actions.

  • Ensure API keys for chosen LLM providers are properly configured via environment variables or the server configuration file.

  • Utilize the headless mode for performance or toggle it to off to observe interactions in real-time for troubleshooting.

  • The server supports persistent logins and cookies via browser profile management, making it suitable for authenticated research tasks.

  • Inputs typically include task descriptions, target URLs, and parameter constraints, while outputs are structured data, research reports, or confirmation of action success.

  • Note that this server requires Playwright as a dependency; it provides utilities to automate the installation of necessary browser drivers.

Repository Stats

Stars
929
Forks
111
Open Issues
11
Language
Python
Default Branch
main
Sync Status
Idle
Last Synced
May 1, 2026, 09:15 AM
View on GitHub