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