nolabs-ai/nono
Sandbox any AI agent in seconds - zero setup, zero latency.
What it solves
nono provides a secure, zero-latency sandbox for running AI agents (such as Claude Code, Codex, and CoPilot) without the need for complex setup, daemons, containers, or virtual machines. It prevents agents from accessing sensitive data like SSH keys and cloud credentials by enforcing a least-privilege security model.
How it works
The tool creates an isolated environment where agents run with restricted access to the filesystem, network, and credentials. It uses a composable JSON policy system to define exactly what an agent can and cannot do.
Crucially, nono implements "sandboxed tool execution," meaning that when an agent calls a tool (like git or gh), that tool is launched in its own separate child sandbox with its own specific policy. This prevents the agent from using a tool to bypass security restrictions or leak secrets.
Who it’s for
- AI Agent Developers: Those who want to publish secure, pre-configured agent packages via the nono registry.
- Engineers and Teams: Professionals who need to run AI agents in production or as part of their local development workflows while maintaining strict security boundaries.
Highlights
- Zero Setup: Runs in seconds with no VM or container overhead.
- Granular Tool Isolation: Isolates delegated tools in child sandboxes with separate filesystem and network rules.
- L7 Filtering: Supports fine-grained network access control, such as restricting a tool to specific API methods and paths.
- Credential Proxying: Injects credentials into tools securely via a proxy rather than giving the agent broad access to raw keys.
- Cross-Platform: Supports macOS, Linux, and Windows (WSL2).
- Registry Support: Access to a community registry of secured agent profiles.