Light weight Code Execution sandbox for Raspberry Pi 4

Описание: Building AI agents that need to test code safely? Here's an approach I've been working on.

So I built an open-source MCP server that uses Docker containers to give agents isolated environments for code testing. In my case, I have deployed the sandbox MCP in my Raspberry pi 4. The key idea is simple:

*What it enables:*
Agents can create temporary Docker containers as sandboxes
Write and execute Python code in isolation
Install packages as needed for testing
Decide whether to deploy successful code or clean up

*Technical approach:*
Uses Docker's containerization for security
MCP protocol for easy agent integration
REST API for direct service access
LangGraph compatible out of the box

The interesting part is watching agents iterate: they create a sandbox, test their code, learn from the results, and either move forward with successful implementations or clean up and try a different approach.

If you're working with AI agents that need code execution capabilities, this might be a useful reference architecture.

Repository: https://github.com/Bkbest/pandora

#AI hashtag#MCP hashtag#Docker hashtag#CodeExecution hashtag#OpenSource hashtag#SoftwareArchitecture hashtag#Agents hashtag#Raspberrypi hashtag#RaspberrypiOS hashtag#LangGraph hashtag#LangChain hashtag#MCP hashtag#FastMCP