Getting started

Three pages, in order. They take you from not having installed anything to a run you can inspect afterwards.

• Introduction — what open-multi-agent is, how goal-first orchestration differs from wiring a graph by hand, and when the difference matters.
• Quick start — a real task DAG in about five minutes, no API key needed, then the one-line switch to a real model.
• Choose a run mode — runAgent, runTeam, or runTasks, decided by who owns the plan and how predictable the graph has to be.

After the first run

• Guides — the decisions that come up once a run is real: orchestration controls, cost and budget limits, and what to wire up before production.
• Reference — the runtime surface page by page: providers and tools, execution and scheduling, observability, evaluation, and the oma CLI.
• Examples — runnable recipes grouped by goal, each with its source.