Brainis has three distinct concepts that people lump together as "agents". They answer different questions.
AI Agents: workers you define
An AI Agent is a worker you build (or install from the marketplace) with a purpose, instructions, tools, and optionally a schedule or trigger. "Watch inbound leads and draft qualification notes." "Summarize every closed-lost deal on Fridays." They are yours: you decide what they do and how much rope they get. See AI Agents.
AI Managers: built-in department heads
AI Managers ship with Brainis, one per operating area - a Product Manager AI, a Project Manager AI, an Account and Delivery AI, a Cortex Executive AI, and an Employee Copilot. They continuously watch their domain, produce reviews and recommendations, and propose actions within their authority. You configure them; you do not build them. See AI Managers.
The Cortex Fleet: the record
The Fleet is not a worker at all - it is the log. Every run by every agent, manager, and Autopilot rule appears there, with the ability to replay a run step by step from its audit trail. When you wonder "what has the AI actually been doing", the Fleet is the answer. See The Fleet.
One governance model
All three operate under the same rules: actions flow through the gateway, autonomy levels and authority contracts bound what runs without approval, and everything lands in the audit chain. Learn the model once in Autonomy levels and it applies everywhere.
Tip: Start with AI Managers (they work out of the box), graduate to marketplace agents, and build custom agents once you have a workflow no template covers.