Grasp (Browser Automation)

🚀 Grasp Project Introduction

Grasp is an open-source and self-hostable intelligent browser agent. It comes with built-in support for MCP and A2A, enabling seamless integration with any other AI applications or agents.

✨ Features

• Isolated Browser Environment: Grasp uses containerization technology to ensure that its operation doesn't interfere with your local working environment. You can easily deploy it on any device or in the cloud.
• Human-like Automated Operations: Grasp operates the browser just like a real person. You can even provide login credentials to allow the agent to access exclusive content on your behalf.
• Natural Language Control: No need to write code. Simply describe your requirements in simple Chinese, and the agent will automatically execute relevant tasks.

🚀 Quick Start

🐳 Prerequisites: Make sure you have Docker installed and running.

Here are the commands to pull and start Grasp:

# Pull the image
docker pull getgrasp/grasp

# Start the container
docker run -d \
  --name grasp-agent \
  -p 3000:3000 \
  -e ANTHROPIC_API_KEY=YOUR_ANTHROPIC_KEY \
  getgrasp/grasp

Replace YOUR_ANTHROPIC_KEY with your actual Anthropic API Key.

After starting, open your browser and visit http://localhost:3000 to access the Grasp console.

📈 Project Updates

Grasp is evolving rapidly. We roll out new features, expand integrations, and optimize the agent experience on a weekly basis. ⭐ Star the repository to stay updated and support the project!

📚 Documentation

• Connect Claude Desktop to Grasp: Use Grasp as a local MCP tool server and enable Claude to perform browser operations.
• Integrate Agents into Grasp: Achieve browser automation collaboration between agents through the A2A protocol.

👥 Community

We look forward to your participation! Join our Discord to share ideas, ask questions, or just hang out.

📄 License

The Grasp project is open-sourced under the BSL-1.1 license.