parallel-subagent-orchestration

You can launch and coordinate multiple specialized Claude agents in parallel to maximize throughput on independent subtasks, delivering 3-5x speedups for concurrent codebase analysis. All selected agents are launched in a single message to enable true parallelism and faster results.

Parallel agent orchestration supports multiple agent types including Explore, general-purpose, Bash, and Plan agents. It provides clear guidance for when to use each type, allowing you to match the agent to the specific concurrent analysis, benchmarking, or documentation task.

You need parallel agent orchestration when your workflow automation involves concurrent analysis, benchmarking, documentation, and reporting across multiple codebases or datasets. It is designed for independent subtasks where launching parallel agents yields a 3-5x speedup.

To synthesize outputs from concurrent agents across multiple codebases, the system applies structured guidelines for prompt construction, independent task prompts, and results synthesis. This enables rapid combination of outputs from concurrent tasks into a unified report.

Yes, parallel agent orchestration is explicitly designed for benchmarking and documentation tasks across multiple codebases. By launching all selected agents in a single message, it maximizes throughput and delivers faster results for these independent subtasks.

Parallel agent coordination is limited to independent subtasks; it is not suitable for sequential or interdependent task coordination where one step relies on the immediate output of another. It requires structured prompt construction to ensure accurate results synthesis across concurrent agents.