Multi-agent orchestration — multiple AI agents collaborating on tasks with division of labor, shared context, and coordination protocols.
Why Evaluating
Relevant to both personal productivity (parallel Claude Code sessions, OpenClaw + Claude Code interplay) and business deployment. The “do more with less” thesis depends on understanding where swarms beat single agents.
Key Questions
- What orchestration patterns exist? (hierarchical, flat, market-based)
- How do swarms handle shared state and context?
- What’s the failure mode when agents conflict?
- Where does this beat single-agent + tools?
- What’s production-ready vs research-only?
Sources to Review
- Claire Vo multi-agent analysis
- Boris Cherny’s parallel Claude pattern (5 terminals + web sessions)
- OpenAI Swarm framework
- CrewAI, AutoGen, LangGraph
part of tooling