Agent Sandbox

Active
GitHub Python Apache-2.0

Description

All-in-One Sandbox for AI Agents that combines Browser, Shell, File, MCP and VSCode Server in a single Docker container, providing a secure isolated execution environment for agents.

Key Features

  • All-in-one sandbox combining Browser, Shell, File, VSCode Server, Jupyter, and MCP in one container
  • Unified filesystem enabling seamless data flow between browser downloads and shell operations
  • Full browser automation via VNC, CDP, and MCP interfaces
  • Pre-configured MCP servers for browser, file, shell, and document processing
  • Multi-language SDK support: Python, TypeScript/JavaScript, and Go
  • 30-second Docker setup with zero configuration required

Use Cases

πŸ’‘ Secure code execution sandbox for AI agent tool calls
πŸ’‘ Browser automation and web scraping with file persistence
πŸ’‘ Development environments with VSCode Server and Jupyter Notebook
πŸ’‘ MCP-compatible agent workflows with sandboxed operations
πŸ’‘ Automated testing and QA with isolated browser and file systems

Tags

automation devops docker agent tools

Categories

πŸ”„ Workflow πŸ—οΈ Sandbox & Execution

Quick Start

Run: docker run --security-opt seccomp=unconfined --rm -it -p 8080:8080 ghcr.io/agent-infra/sandbox:latest. Access: http://localhost:8080/v1/docs (API docs), http://localhost:8080/vnc/ (browser), http://localhost:8080/code-server/ (IDE). Python SDK: pip install agent-sandbox.
Visit GitHub Visit Website

Related Projects