langchain-ai

langchain-ai / openshell-deepagent

Public

A general-purpose coding agent that runs inside an NVIDIA OpenShell sandbox, orchestrated by Deep Agents and powered by NVIDIA Nemotron. The agent writes and executes code in an isolated, policy-governed Linux environment.

47
8
100% credibility
Found Mar 19, 2026 at 47 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
Python
AI Summary

This project creates a local AI agent for coding and analysis that runs code in a secure, policy-controlled sandbox on your machine using NVIDIA's AI model.

How It Works

1
🕵️ Discover the secure coding helper

You hear about a smart AI assistant that writes and runs code safely right on your own computer, without needing the internet for everything.

2
🔑 Get your AI thinking pass

Sign up for a free pass from NVIDIA so your assistant can use their powerful brain.

3
📦 Prepare your computer

Install a few simple tools like a package helper and make sure your container app is running.

4
🚀 Create a safe playground

Start a protected space where the AI can play with code without causing any harm.

5
🎉 Wake up your coding wizard

Launch the assistant with one command and open its friendly web chat interface.

6
💬 Give it tasks to solve

Chat with your AI like a friend, asking it to run commands, write files, or analyze data safely.

✅ Enjoy secure results

Watch your assistant complete coding jobs perfectly, with built-in rules keeping everything safe and under control.

Sign up to see the full architecture

5 more

Sign Up Free

Star Growth

See how this repo grew from 47 to 47 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 openshell-deepagent?

OpenShell DeepAgent is a Python-based general-purpose coding agent that executes code inside an isolated, policy-governed Linux environment powered by NVIDIA OpenShell, Deep Agents orchestration, and NVIDIA Nemotron models. It lets agents write, run, and iterate on code—like Python scripts or bash commands—in a secure on-prem sandbox without cloud dependencies, enforcing strict controls on filesystem, network, and processes. Developers get a drop-in backend for LangChain-style agents, with persistent local storage for memory and skills.

Why is it gaining traction?

It stands out by combining Deep Agents' task planning with OpenShell's Docker-like security that actually blocks malicious actions, like outbound network calls to evil endpoints, no matter the prompt. Setup is straightforward via uv sync, Docker, and a free NVIDIA API key, then spin up a persistent sandbox and run via LangGraph dev server for instant Studio UI access. The policy YAML lets you tweak permissions on the fly, making it a secure alternative to cloud sandboxes for general-purpose coding agents.

Who should use this?

AI engineers building multi-step agents that need to run untrusted code locally, like data analysis or script generation without risking your host machine. LangChain users swapping in OpenShell for safer execution in tools like execute or write_file. Teams prototyping general-purpose coding in isolated Linux environments before production.

Verdict

Worth a spin for secure, local agent experimentation—docs and smoke tests are solid—but with 47 stars and 1.0% credibility score, it's early-stage and lacks broad testing. Try the file roundtrip or stats script demos if on-prem Nemotron appeals.

(198 words)

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.