TrendWhat rose on GitHub, every day at 07:30 KST · 한국어

GitHub · as of November 15, 2026 · View on GitHub →

NVIDIA/OpenShell

AI agent runtime

+371 stars in 7 days · trending 12 of the last 30 days

Secure AI agent file access and API calls with policy-driven enforcement

When an AI agent opens files, installs packages, or calls APIs, your data and secret keys may be at risk. OpenShell provides an execution environment that allows agents to use necessary features while strictly blocking access to restricted areas. You declare what an agent can do through policies, and the system enforces them.

CNCF migration and 0.1.x changes

OpenShell is now part of the CNCF (Cloud Native Computing Foundation), with development moving to ai-openshell and governance shifting to the CNCF. Until the migration is complete, the code remains managed in the NVIDIA/OpenShell repository. Version 0.1.x introduces a stable release cycle, new isolation primitives, an extended extension surface, and new APIs. It is recommended to check the upgrade guide to understand the changes.

Kernel-level enforcement and schema validation

OpenShell controls agent behavior in two ways. First, it instruments the kernel to enforce policies at runtime for all file access, system calls, and network connections. Each agent runs in an isolated sandbox where actual credentials are invisible and are only added to requests for approved endpoints. Second, it performs schema validation before applying policy changes. This detects risky access, such as new host access or API calls, in advance and holds them for human review.

Installation and first run

You need a Linux, Apple Silicon-based macOS, or Windows environment with WSL 2 (experimental). Docker, Podman, or host virtualization is also required. Running the installation script sets up the CLI and local gateway. The default sandbox image is a minimal Ubuntu without an agent. To run an actual agent, follow the guide for using OpenCode with a free OpenRouter model, which demonstrates the process of approving new access requests from the agent.

Agent skills and SDK

You can install public skills that help coding agents drive the OpenShell CLI, write sandbox policies, and debug the gateway. Install them with the npx skills add NVIDIA/OpenShell command, which works without a source checkout. You can connect applications to the OpenShell gateway using Python, TypeScript, Go, and Rust SDKs. The SDKs do not install the CLI, and it is recommended to use the same OpenShell release for both the SDK and the gateway where possible.

Pre-use considerations

OpenShell is licensed under the Apache License 2.0, and the software is provided 'AS IS'. It automatically retrieves and interacts with external materials, but users must verify the security and appropriateness of those materials themselves. It collects anonymous telemetry but does not collect sandbox names, file paths, prompts, credentials, or user content. You can disable or remove telemetry at compile time. For Kubernetes deployments, the CNI must enforce NetworkPolicy, and Windows native support is provided through Microsoft MXC, so you should check current limitations.

Written by AI from this repository's README on November 10, 2026. For the facts, the original on GitHub is the reference.

By the numbers

GitHub trending history
30 days between Jun 19, when Trend began recording, and Nov 15, 2026 · 12 of the last 30 days · first on Sep 29 · best daily rank #1 (Sep 30)

Read from GitHub, as of November 15, 2026.

Language
Rust
Stars
21,647
Stars gained in 7 days
+371
Contributors
143
Forks
2,115
Latest release
v0.1.8 · November 10, 2026
Commits
2,044
Last commit
November 13, 2026
Open issues
463
Open pull requests
237

View on GitHub → · Homepage

README badge

trend.ondo.sh · GitHub daily trending #1 · September 30, 2026

The badge above shows the highest this repository has ranked on GitHub's daily trending since June 19, 2026, and the day it did. Paste the Markdown below into your README. The badge opens this page.

[![trend.ondo.sh](https://trend.ondo.sh/github/r/NVIDIA--OpenShell/badge.en.svg)](https://trend.ondo.sh/github/r/NVIDIA--OpenShell/en.html)