Concepts

What is multi-agent orchestration?

Multi-agent orchestration is coordinating several AI agents so they hand off to each other to complete a job that one agent shouldn’t do alone. Each agent has a focused role; the orchestration decides the order and how results flow between them.

Orchestration modes in AgentMesh

1. Sequential — one agent after another, passing results along.
2. Parallel — several agents at once, for speed.
3. Conditional — branch based on a result (if/else).
4. Fan-out — split work across many agents, then combine.
5. DAG (dependency graph) — steps run when their dependencies finish.
6. Loop — repeat the steps until a condition or a max number of iterations.

Why it matters

Focused agents are easier to build, test, and trust than one giant prompt. Orchestration lets you compose them — e.g. a triage agent routes to a specialist agent, which drafts a reply a review agent checks.