Orchestration
AI agent orchestration is the coordination of several agents and tools so they work through one job in the right order. It decides which agent acts next, what it receives, and when the job is done.
In a run
In the sandbox, a workflow-builder agent opens a session for the job and issues the credential for it. The other agents do the steps. When the job is done, the credential is closed.
Our word for it
The workflow, with defined steps.
Coordination is not control
A coordinator that passes work between agents decides the order. It does not, by itself, decide what each agent may do.
Ares adds that part. The policy is checked on every call. The permission narrows at each handoff. A step outside the credential is refused, and the reason goes on the record.
In the supplier follow-up, the credential names five actions, among them reading open orders, emailing the supplier on file and changing a delivery date with buyer approval. A request to send a drawing to a new address is not on that list, so it is refused.
The framework you already use
Ares does not depend on the agent framework. Calls on every path get the same policy check and the same record: see A2A for how an agent connects. If your team already coordinates its agents one way, it keeps doing so.
Where it stops
Ares runs the steps you defined and checks each call. It does not make a model plan better. It does not cover agents you have not enrolled.
Run the supplier follow-up to watch one job cross four systems.
Related terms
- Multi-agent workflowA multi-agent workflow is a job split across several AI agents, each doing one part and handing work to the next.
- DelegationDelegation is one party handing a task, and part of its authority, to another.
Tell us about one workflow.
We will say whether an agent should do it, and what the first step costs.
A senior member of the team replies within two business days.