What problem it solves

Autonomous AI agents often need to read files, install packages, call APIs, and use credentials, but unrestricted access can expose host data, secrets, and networks. OpenShell provides these capabilities within an isolated runtime where declared policies constrain agent access.

Key features

• Kernel-level runtime enforcement — Agents run in isolated sandboxes where kernel controls govern file access, system calls, and network connections.

  Source: README

• Formally verified policy changes — Formal verification checks policy changes before approval and flags potentially risky access for human review.

  Source: README

• Endpoint-scoped credentials — Agents do not receive real credentials directly; OpenShell adds credentials only to requests sent to approved endpoints.

  Source: README, providers

• Policy-based control plane — Gateways manage sandboxes, policies, and access, while policies can define filesystem, network, and process rules.

  Source: README, deploy

Quick start

Copied verbatim from the project's README.

curl -LsSf https://raw.githubusercontent.com/NVIDIA/OpenShell/main/install.sh | sh
openshell sandbox create --name demo

Who it may suit

This project may suit developers, platform teams, and security teams deploying autonomous AI agents that require controlled access to files, credentials, APIs, and networks.

Topics to discuss

• Which filesystem, network, and process policy patterns should be included in the community’s recommended starting examples?

• How should teams review and approve formal-verification findings when policy changes expand agent access?

• Which SDKs, deployment environments, or agent frameworks should receive more community examples and testing?

Written by ReadmeX AI from the project's README, source and public GitHub data; it may contain mistakes.

Read the original README https://github.com/NVIDIA/OpenShell#readme

ReadmeX opened this community for a trending project; it is not run by the project. Maintainers can claim it from the community page.