task-router

Task routing analyzes your workload and context to recommend the optimal agent, defaulting to Sonnet when uncertain to minimize expensive Opus usage. It matches simple tasks to cheaper alternatives like Haiku before you spawn sub-agents.

Multi-agent task routing is the process of analyzing tasks to deterministically select the best AI model, such as Haiku or Codex, before execution. You need it when balancing cost and performance across varying workloads to prevent defaulting to expensive agents.

You integrate agent selection by passing task and context data to the router, which returns a structured output containing the recommended agent and reasoning. Use this structured output directly before your sessions_spawn call to route tasks efficiently.

Yes, task routing covers simple to complex workloads using a heuristic router. It evaluates the specific context of your task to match it with the appropriate agent, ensuring cost optimization across all levels of complexity.

No, the router operates independently without any specific dependencies. It analyzes tasks and context directly to provide actionable agent recommendations and reasoning for your session workflows.