nono
ActiveDescription
A capability-based, multiplexing sandbox tool built for developers — run agents securely without needing any additional infra, zero setup, zero latency.
Key Features
- Capability-based, policy-governed runtime for AI agents
- Landlock (Linux) + Seatbelt (macOS) kernel-level sandbox isolation
- Credential proxy injection mode, API keys never enter the sandbox
- Sigstore signing and verification of instruction file integrity
- Multiplexing for parallel agent sandboxes
- Policies are composable, auditable, and versionable
Use Cases
💡 Safely running untrusted AI agent code
💡 Enterprise-level agent policy governance and audit compliance
💡 Isolated agent task execution in CI/CD pipelines
💡 Local secure debugging of agent applications for developers
Categories
Quick Start
brew install nono, then run nono run --profile always-further/claude -- claude to launch a sandboxed agent.