Documentation

ColleagueOne gives organisations a structured way to put AI colleagues to work. These guides explain the full loop: define a goal, review the colleague, let it work in an isolated cloud computer, and review the result.

Choose a path

• New to ColleagueOne? Start with Getting started and Core concepts.
• Creating a role? Use the Colleague builder, then configure Skills and plugins and Connections.
• Reviewing work? Learn the Workspace, Approvals and Memory.
• Running ColleagueOne for an organisation? See Admin and members, SSO and SCIM and the Security model.
• Integrating through code? Begin with the API quickstart, then use the generated API reference.

The core loop

1. Describe the goal. State the outcome, inputs and review expectations.
2. Review the proposed colleague. Check its persona, tools, rules, model and schedule.
3. Run the work. The colleague works inside its own Firecracker microVM sandbox.
4. Review the result. Outputs appear in the workspace; questions and sensitive actions appear in Needs You.

Documentation principles

These pages distinguish between colleague capability and human authority. A result is something to inspect, not an automatic statement of truth. An approval is a deliberate decision, not a notification to click through. Memory is reviewable information, not an invisible permanent profile.

ColleagueOne is currently an invite-only beta. Interfaces and documented behavior may evolve before general availability.