orchestration

Multi-agent workflows across Claude Code agents are coordinated using orchestration patterns like Fan-Out to spawn multiple subagents for parallel work and independent results. This enables concurrent research and implementation across decomposed domains.

Aggregating outputs across sequential agents is handled through Pipeline and Map-Reduce orchestration patterns. These sequence and aggregate outputs across stages, allowing subagents to process data step-by-step and synthesize final results.

Team-based coordination supports long-running and inter-dependent tasks across Claude Code agents. It applies domain decomposition with clear guardrails, allowing multiple agents to sustain complex workflows that require ongoing interaction.

Multi-agent orchestration is needed for complex software engineering tasks requiring parallel research, planning, and synthesis. It satisfies orchestration patterns and domain decomposition when a single agent cannot efficiently handle the workload.

Designing a multi-agent plan involves structuring four agents to coordinate feature research and implementation using subagents and teams. The orchestration skill provides reference-guided patterns to decompose the domain and synthesize results effectively.