chu2bard

chu2bard / execbox

Public

Code execution sandbox for AI agents with safety controls

agents code-execution python safety sandbox
19
0
89% credibility
Found Feb 11, 2026 at 19 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
Python
AI Summary

Execbox is a tool that creates a secure, isolated space for running code snippets, especially those generated by AI, with built-in limits on resources and permissions to prevent harm.

How It Works

1
🔍 Discover safe code runner

You hear about execbox, a helpful tool that lets AI assistants try out code ideas without any danger to your computer.

2
📦 Bring it into your project

You add execbox to your setup, making it ready to protect and run code snippets.

3
🛡️ Set your safety boundaries

You pick simple rules like time limits, memory caps, and blocks on risky actions to keep everything secure.

4
▶️ Run the code safely

You give a piece of code to execbox, and it handles the execution in a fully protected space.

5
📊 Review the results

You receive the output, any messages, and confirmation of success or issues, all neatly captured.

✅ AI experiments securely

Your AI agent has safely tested its code ideas, giving you reliable results with zero risk.

Sign up to see the full architecture

4 more

Sign Up Free

Star Growth

See how this repo grew from 19 to 19 stars Sign Up Free
Repurpose This Repo

Repurpose is a Pro feature

Generate ready-to-use prompts for X threads, LinkedIn posts, blog posts, YouTube scripts, and more -- with full repo context baked in.

Unlock Repurpose
AI-Generated Review

What is execbox?

Execbox is a Python library for safely running untrusted code snippets in isolated subprocesses, designed for AI agents that need code execution without host risks. It enforces resource limits on timeout, memory, CPU time, and output size, plus policies blocking dangerous imports, builtins, and network access to prevent code execution vulnerabilities. Users get structured results with stdout, stderr, status, and elapsed time via a simple API.

Why is it gaining traction?

Execbox hooks developers building agents by combining easy Python code execution with MCP-style safety—like denying os or subprocess—that lets code execution proceed securely from models like Gemini or Anthropic tools. It cuts code execution time python for efficient agent loops, standing out from verbose sandboxes by prioritizing agent-friendly defaults over complex setups. In code github ai circles, its lightweight policies enable safer code execution with pre-trained language models.

Who should use this?

AI agent builders integrating code interpreters for tasks like data processing or simulations. LLM devs creating code execution ai studios or code execution gemini plugins. Backend teams using code execution with mcp anthropic to harden agents against exploits.

Verdict

Solid prototype pick for AI experimentation, scoring 0.8999999761581421% credibility with 19 stars and MIT license, but basic docs and no visible tests signal early maturity. Use for non-prod agent sandboxes; watch for hardening before scaling.

(198 words)

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.