All skills
Skillintermediate

Use Case: Multi-Agent Research System

<div align="center">

Claude Code Knowledge Pack7/10/2026

Overview

<div align="center">

šŸ  Home › šŸ“˜ Guides › šŸŽÆ Use Cases › Multi-Agent Research

← Use Cases ā”ā”ā—ā”ā”ā”ā”ā”ā”ā”ā”ā”ā”ā”ā”ā”ā”ā”ā”ā”ā”ā”ā”ā”ā”ā”ā”ā”ā”ā”ā”ā”ā”ā”ā”ā” Production Code Review →

</div>

Use Case: Multi-Agent Research System

Source: Anthropic Engineering Blog - June 2025


Problem

Synthesizing comprehensive research from multiple sources requires:

  • Parallel information gathering
  • Domain specialization
  • Quality synthesis

Solution Architecture

%%{init: {'theme': 'base', 'themeVariables': {'lineColor': '#64748b'}}}%%
flowchart TB
    classDef main fill:#8b5cf6,stroke:#7c3aed,stroke-width:2px,color:#ffffff
    classDef subagent fill:#ec4899,stroke:#db2777,stroke-width:2px,color:#ffffff
    classDef tool fill:#f59e0b,stroke:#d97706,stroke-width:2px,color:#ffffff
    classDef state fill:#10b981,stroke:#059669,stroke-width:2px,color:#ffffff

    USER["šŸ™‹ā€ā™€ļøšŸ“„ Research Query"] --> LEAD["šŸ” Main Agent"]:::main

    LEAD -->|"🪺 Task"| PLAN["Plan research strategy"]
    PLAN --> SPAWN["Spawn specialized researchers"]

    subgraph PARALLEL["šŸš‚ Parallel Execution"]
        R1["🐦 Academic Researcher"]:::subagent
        R2["🐦 Industry Researcher"]:::subagent
        R3["🐦 News Researcher"]:::subagent
    end

    SPAWN --> R1 & R2 & R3

    R1 --> T1["šŸ”Œ Perplexity"]:::tool
    R2 --> T2["šŸ”Œ Firecrawl"]:::tool
    R3 --> T3["šŸ”Œ WebSearch"]:::tool

    T1 & T2 & T3 --> COLLECT["šŸ” Main Agent collects"]:::main
    COLLECT --> SYNTH["🐦 Synthesis Agent"]:::subagent
    SYNTH --> REPORT["āœ… Final Report"]:::state

Patterns Used

PatternRole
šŸ¦‘ Orchestrator-WorkersšŸ” Main Agent spawns specialized researchers
šŸš‚ Parallel Tool CallingMultiple researchers work simultaneously
🧬 Master-CloneEach researcher has isolated context

Implementation

šŸ” Main Agent Prompt

main_agent_prompt = """
You coordinate research by:
1. Breaking query into research domains
2. Spawning domain-specific researchers
3. Collecting and synthesizing results
"""

🐦 Researcher Subagent

researcher_prompt = """
You research {domain} using available tools.
Return structured findings with citations.
"""

Subagent Definition

# .claude/agents/academic-researcher.md
---
name: academic-researcher
description: Researches academic papers and scholarly sources
tools: Read, WebFetch, WebSearch
model: sonnet
---

You are an academic research specialist. Your task is to:
1. Search for relevant academic papers
2. Extract key findings and citations
3. Return structured research notes

Focus on peer-reviewed sources and cite properly.

Why This Works

AspectBenefit
SpecializationEach subagent focuses on one domain
ParallelismIndependent searches run concurrently
IsolationSubagents don't pollute each other's context
SynthesisšŸ” Main Agent has full picture for final output

Flow Summary

šŸ™‹ā€ā™€ļøšŸ“„ Research Query
    │
    ā–¼
šŸ” Main Agent (Plan strategy)
    │
    ā”œā”€ā”€ā–ŗ 🐦 Academic Researcher ──► šŸ”Œ Perplexity
    ā”œā”€ā”€ā–ŗ 🐦 Industry Researcher ──► šŸ”Œ Firecrawl
    └──► 🐦 News Researcher ──► šŸ”Œ WebSearch
            │
            ā–¼
    šŸ” Main Agent (Collect results)
            │
            ā–¼
    🐦 Synthesis Agent
            │
            ā–¼
    āœ… Final Report

<div align="center">

← Use Cases ā”ā”ā—ā”ā”ā”ā”ā”ā”ā”ā”ā”ā”ā”ā”ā”ā”ā”ā”ā”ā”ā”ā”ā”ā”ā”ā”ā”ā”ā”ā”ā”ā”ā”ā”ā” Production Code Review →

</div>