[ ABORT TO HUD ]
SEQ. 1
SEQ. 2
SEQ. 3
SEQ. 4
SEQ. 5
SEQ. 6
SEQ. 7
SEQ. 8

Building a Multi-Agent Pipeline

👥 Multi-Agent Systems⏱ 12 min⭐ 100 BASE XP

Hands-On: Research-Write-Review Pipeline

Let's build a practical 3-agent pipeline: Researcher gathers information, Writer drafts content, Reviewer provides feedback. The loop continues until quality is sufficient.

System Architecture

┌─────────────────────────────────────┐
│           ORCHESTRATOR               │
│  (manages handoffs, tracks quality)  │
├──────┬──────────┬──────────┬────────┤
│ Step │  Agent   │  Input   │ Output │
├──────┼──────────┼──────────┼────────┤
│  1   │Researcher│ Topic    │ Notes  │
│  2   │ Writer   │ Notes    │ Draft  │
│  3   │ Reviewer │ Draft    │ Score  │
│  4   │ Writer   │ Feedback │ v2     │
│  ...repeat until score >= 8/10...   │
└─────────────────────────────────────┘

Agent Definitions

const agents = {
  researcher: {
    system: "You are a research specialist. Given a topic, search the web and compile a structured research brief with key facts, statistics, and expert opinions. Output JSON.",
    tools: ["web_search", "read_url"],
    model: "claude-sonnet-5"
  },
  writer: {
    system: "You are a technical writer. Given research notes (and optional reviewer feedback), write a clear, engaging blog post. Use examples and code snippets.",
    tools: [],
    model: "claude-sonnet-5"
  },
  reviewer: {
    system: "You are an editor. Score the draft 1-10 on accuracy, clarity, and engagement. Provide specific, actionable feedback. Output JSON: {score, feedback[]}",
    tools: [],
    model: "claude-haiku-4-5" // Cheap model for review
  }
};

Key Implementation Tips

TipWhy
Use a cheap model for the reviewerReview doesn't need creativity, saves 5-10x on tokens
Cap iterations at 3Diminishing returns after 2-3 revision cycles
Pass summaries, not full outputsWriter only needs the feedback, not the full review analysis
Log every handoffEssential for debugging - you need to see what each agent received
💡 Key Insight: The orchestrator is the most important component. It decides when to move to the next agent, when to loop, and when to stop. A well-designed orchestrator with mediocre agents outperforms mediocre orchestration with perfect agents.
SYNAPSE VERIFICATION
QUERY 1 // 2
In a multi-agent pipeline, which component is most critical to overall quality?
The writer agent's prompt
The orchestrator that manages handoffs, quality gates, and iteration
The most expensive model
The number of agents