AI agents · Types
What is an agent team?
The short answer
Specialists hand work directly to one another, and whichever holds the work deals with the user. No single agent is in charge.
It is one of the 4 forms of multi-agent system, with the agent pipeline and the manager agent and the guardian agent.
| What starts it | Any of the above |
|---|---|
| Who decides the next step | Whichever agent holds the work |
When should you use an agent team?
Use it when a case moves between specialisms and each specialist should own its part.
When should you not?
Not when you need one accountable point of control. This is the least mature form. One analyst rates it the highest risk of four patterns, and it is the hardest to trace.
What does it need to run safely?
It needs rules on who may hand to whom, detection of work going round in circles, and one record across every handover.
An example
An intake agent passes a customer to the billing agent, which passes a cancellation to the retention agent.
What else is it called?
Handoff, swarm, group chat.
How do you build one on Katonic?
- New agent, Coordinate a team. Start blank. The sections are Identity, Specialists, Routing, Guardrails, Advanced.
- Identity. Name the team and pick the Owner team.
- Specialists. Attach the agents this coordinator routes to. The page says: "Pick at least 2", and it will not publish with fewer.
- Routing. Keep LLM router. The Router prompt is written from the specialists list; edit it to sharpen who gets what.
- Test. The tab holds routing test cases, each with an expected specialist. Choose Run all and read the routing accuracy and the confusion matrix.
- Deploy, then Publish.
Questions people ask
What starts an agent team?
Any of the above.
Who decides its next step?
Whichever agent holds the work.
What is a multi-agent system?
A multi-agent system is several agents, each with its own instructions and tools, working on one piece of work. The forms differ in who decides the order. OpenAI advises starting with one agent and moving to several only when needed. Microsoft's guidance is to start with several where the work crosses security boundaries or teams, and otherwise to test a single agent first. Gartner expects 70 percent of AI applications to use multi-agent systems by 2028.