OneCLI is an open-source sandboxed agent harness that provisions and manages AI agents for teams.
The tool addresses the challenge of running autonomous agents safely across organizations. Most AI agents are built for single users, but deploying them at team scale requires managing secrets, enforcing permissions, and tracking which agent belongs to whom. OneCLI solves this by creating one sandboxed agent per employee, routed through a central gateway that injects credentials and enforces team policies. Agents are durable entities with persistent state rather than single-prompt interactions, and they can be accessed from a dashboard or Slack. The platform integrates with identity providers to provision agents tied to employee identities, and it supports deterministic human-in-the-loop approvals for sensitive actions like sending emails or deleting resources.
Teams should adopt OneCLI if they run autonomous agents like Hermes, OpenClaw, or NanoClaw and need centralized credential management and policy enforcement across multiple users. It suits organizations that want to give employees AI agents capable of real work while maintaining security boundaries and audit trails. The tool is available both as a cloud-hosted service and for self-hosting.
Development activity shows consistent engagement with the codebase through regular commits and pull request activity. The project maintains active communication channels with a Discord community for user support and feedback. The maintainers have demonstrated responsiveness to issues and pull requests, indicating ongoing stewardship of the platform.
