cookbooks

Multi-Agent Pipeline

An orchestrator analyzes the task and delegates to a researcher (web search) and a writer (document creation), all running on a shared Docker bridge network.

agentfile.yaml

agents:
  orchestrator:
    metadata:
      role: Task coordinator
      goal: Break down content creation tasks and delegate to specialists
      instructions: |
        You coordinate content creation. Given a topic:
        1. Delegate research to the researcher agent
        2. Pass findings to the writer agent with specific instructions
        3. Review the final output and request revisions if needed
    runtime:
      provider: anthropic
      model: claude-sonnet-4-6
    collaborators: [researcher, writer]

  researcher:
    metadata:
      role: Web researcher
      goal: Find accurate, comprehensive information on any topic
      instructions: Return structured findings with source URLs.
    runtime:
      provider: anthropic
      model: claude-haiku-4-5-20251001
      temperature: 0.1
    tools:
      - name: web_search
        source: mcp://brave-search
      - name: github
        source: mcp://github

  writer:
    metadata:
      role: Technical writer
      goal: Write clear, well-structured technical documents
      instructions: |
        Write in a professional, concise style. Use markdown.
        Include a TL;DR, structured sections, and a conclusion.
    runtime:
      provider: openai
      model: gpt-4o
      temperature: 0.6
    tools:
      - name: filesystem
        source: mcp://filesystem

Run the pipeline

Bash

ninetrix build
ninetrix up

# Send a task to the orchestrator
ninetrix invoke --message "Write a technical overview of the Model Context Protocol"

# Watch the trace
ninetrix trace
ninetrix logs

#agentfile.yaml

#Run the pipeline

agentfile.yaml

Run the pipeline