parcadei

parcadei / ouros

Public

A sandboxed Python runtime for AI agents, written in Rust.

114
10
100% credibility
Found Feb 20, 2026 at 88 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
Rust
AI Summary

Ouros provides a fast, sandboxed Python runtime in Rust for AI agents to execute generated code safely with persistent sessions, resource limits, and controlled external access.

How It Works

1
👀 Discover safe code running

You hear about a tool that lets AI assistants run Python code securely without risking your computer.

2
📦 Add to your project

You easily add it to your Python setup so your AI can use it right away.

3
🔒 Create a safe playground

You make a protected space where AI code runs without touching your files or network.

4
▶️ Feed in data and run

You give starting values and watch the code compute results safely.

5
Need helpers or ongoing work?
🔧
Add helper actions

Connect safe outside functions like web lookups so AI can do more.

🔄
Start a persistent workspace

Create a notebook-like space where variables stick around for multi-step thinking.

🚀 AI thinks and acts safely

Your assistant now runs complex code turns after turns, rewinds mistakes, and stays secure.

Sign up to see the full architecture

4 more

Sign Up Free

Star Growth

See how this repo grew from 88 to 114 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 ouros?

Ouros delivers a sandboxed Python runtime in Rust for safe execution of AI agent code, blocking filesystem, network, and subprocess access while allowing controlled external functions. Users get stateful REPL sessions for persistent variables, fork/rewind for experimentation, snapshots for serializable pauses, and 72 stdlib modules like json, asyncio, and pathlib. Bindings for Python, JS/TS, and Rust plus an MCP server enable run python sandbox workflows across tools.

Why is it gaining traction?

It boots in under 1 microsecond for agents, outpacing CPython startup, with Ruff-based type checking and hard resource limits on memory, time, and recursion to kill rogue scripts. Unlike full VMs or containers, ouros intact hide offers lightweight sandboxed python execution with snapshot/resume for distributed agent states, hooking devs building recursive language models or tool-calling agents.

Who should use this?

AI engineers crafting agents that generate and run Python code, backend teams needing sandboxed python interpreter for user-submitted scripts in plugins or judges, or RLM implementers handling large data without bloating LLM context. Ideal for Cursor/Claude users via MCP or JS devs wanting safe sandbox python in Node.

Verdict

Early at 19 stars and 1.0% credibility score, but solid docs, tests, and Monty fork make it viable for prototypes; monitor for stability before production agents.

(198 words)

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.