Execute code in isolated Docker containers with support for Python, Node.js, and R, ideal for AI agents.
Sponsored
The Sandbox Agent is a powerful LangGraph agent designed for secure, Docker-based sandboxed code execution. Each session runs in an isolated, hardened Docker container with a persistent kernel, supporting Python (IPython), Node.js, and R. It offers a robust environment for AI agents by providing advanced features like crash detection, provider-agnostic LLM configuration, rich display outputs, and seamless cross-session file transfer. The agent is versatile, available as an interactive CLI, an MCP server for integration with tools like Cursor, a comprehensive REST API, and a React-based frontend, catering to various deployment and interaction needs.
Key Features
01Docker isolation with hardened security (non-root, PID/memory limits, tmpfs)
02Provider-agnostic LLM integration with automatic vision support
030 GitHub stars
04Multi-runtime support for Python, Node.js, and R environments
05Comprehensive file management including upload, export, import, and cross-session transfer
06Persistent state for variables between code executions, like Jupyter cells
Use Cases
01Developing web-based platforms that provide multi-language code environments with file persistence
02Building secure AI agents requiring sandboxed execution for Python, Node.js, or R code
03Integrating interactive code execution capabilities into IDEs or client applications via the MCP server
