About Nono
Nono is ephemeral micro sandboxes for AI agents, designed for developers and teams building AI-powered workflows. It delivers zero-setup, zero-latency isolation for each tool invocation, with auditable, policy-driven controls.
Key Features
Practically, you run a tool invocation, the supervisor enforces argv and credential policies, then a fresh sandbox isolates the tool with access limited to what’s allowed, before the run ends and the sandbox is torn down:
Ephemeral Sandbox per Invocation
A fresh, invocation-scoped sandbox surrounds the tool, with a read-only workspace, no filesystem writes, no host secrets, bounded stdout/stderr, and network access through the Nono proxy.
Policy-Driven Access
The system validates the caller, argv, and capabilities and can pause for human approval if a call falls outside policy.
Credential Safety & Proxyed Network
Real credentials stay with the supervisor; the child only uses a phantom token, while outbound requests go through the TLS-protected proxy to the target API.
Tamper-Evident Audit Trail
Each invocation is hash-chained into a Merkle-root record; the sandbox is destroyed on exit.
Summary
Best for AI/ML teams building agent workflows, platform engineers orchestrating tool runs, and security/compliance teams auditing tool usage.
