Supervisor Multi-Agent Router

A supervisor agent delegates sub-tasks to specialist agents with narrow tools and merges their outputs.

When to use

Only when one agent with many tools gets confused. Each specialist gets a small toolset and a focused prompt; the supervisor sees summaries, not raw tool output.

Steps

1. Define specialists (e.g. researcher, coder, writer) each with its own system prompt and tools.
2. Expose each specialist to the supervisor as a single tool: delegate_to_<name>(task: string).
3. Specialists return a concise result (not their full transcript).
4. Supervisor decides the next delegation or returns the final answer.
5. Enforce a global step and cost budget across all agents.

Definition

supervisor:
  model: strong-model
  prompt: >
    You coordinate specialists. Delegate one clear sub-task at a time.
    Integrate results and answer the user. Do not do specialist work yourself.
  tools: [delegate_to_researcher, delegate_to_coder, delegate_to_writer]
specialists:
  researcher: {model: fast-model, tools: [web_search, fetch_url]}
  coder:      {model: code-model, tools: [read_file, write_file, run_tests]}
  writer:     {model: fast-model, tools: []}
budget: {max_total_steps: 30, max_cost_usd: 1.00}

Failure modes

• Supervisor ping-pongs between workers without progress; cap handoffs.
• Workers lose context the supervisor had; pass explicit task briefs.
• Multi-agent overhead is unnecessary for simple tasks; start with one agent.

Inspiration

Real-world reports that inspired or relate to this blueprint, credited to Agent Use Cases:

• Explore open math problems with a multi-agent system (agentusecases.ai)

Markdown version: /blueprints/supervisor-multi-agent.md

Related blueprints