Skip to main content

​
Parallel Market Analysis with Sub-Agents

This example demonstrates how a single coordinator agent can dynamically create specialized sub-agents and delegate parallel research tasks — perfect for complex analyses that benefit from domain-specific expertise.

​
What This Example Shows

  • Enabling sub-agent delegation with max_loops="auto"
  • How the coordinator autonomously creates and assigns work to sub-agents
  • Parallel execution across multiple research domains
  • Result aggregation into a unified report

​
How Sub-Agents Work

Unlike agent handoffs where specialist agents are pre-defined in your request, sub-agents are created dynamically at runtime by the coordinator. The coordinator decides how many agents to create and what each one specializes in.

​
Step 1: Setup

import requests
import os
import json

API_BASE_URL = "https://api.swarms.world"
API_KEY = os.environ.get("SWARMS_API_KEY", "your_api_key_here")

headers = {
    "x-api-key": API_KEY,
    "Content-Type": "application/json"
}

​
Step 2: Configure the Coordinator Agent

The coordinator agent needs max_loops="auto" to enable autonomous tool use, including create_sub_agent and assign_task: 
def analyze_market(company: str) -> dict:
    """Run a parallel market analysis using sub-agent delegation."""

    payload = {
        "agent_config": {
            "agent_name": "Market-Analysis-Coordinator",
            "description": "Senior market analyst that coordinates parallel research streams",
            "system_prompt": (
                "You are a senior market analyst coordinating a research team. "
                "For each analysis request:\n"
                "1. Create specialized sub-agents for each research domain\n"
                "2. Assign specific research tasks to each sub-agent\n"
                "3. Wait for all results and compile a comprehensive market report\n\n"
                "Always create at least 3 sub-agents covering: "
                "financial analysis, competitive landscape, and industry trends."
            ),
            "model_name": "gpt-4.1",
            "max_loops": "auto",
            "max_tokens": 8192,
            "temperature": 0.3
        },
        "task": (
            f"Conduct a comprehensive market analysis for {company}. "
            "Create specialized sub-agents and delegate the following research areas:\n"
            "1. Financial Analysis: Revenue trends, profitability, key financial metrics\n"
            "2. Competitive Landscape: Major competitors, market share, differentiation\n"
            "3. Industry Trends: Market growth, emerging technologies, regulatory changes\n"
            "4. Risk Assessment: Key risks, vulnerabilities, mitigation strategies\n\n"
            "Compile all findings into a structured executive briefing."
        )
    }

    response = requests.post(
        f"{API_BASE_URL}/v1/agent/completions",
        headers=headers,
        json=payload,
        timeout=300
    )

    return response.json()

​
Step 3: Run the Analysis

result = analyze_market("Tesla Inc.")

if result.get("success"):
    print(f"Agent: {result['name']}")
    print(f"Output:\n{result['outputs'][:2000]}...")
else:
    print(f"Error: {result}")
Expected Behavior: The coordinator will autonomously:
  1. Create sub-agents — e.g., “Financial-Analyst”, “Competition-Researcher”, “Industry-Trends-Analyst”, “Risk-Assessor”
  2. Assign tasks — Each sub-agent receives a focused research task
  3. Execute in parallel — Sub-agents run concurrently
  4. Compile results — The coordinator synthesizes all findings into a unified report
Sub-agent workflows take longer than single-agent calls since multiple agents are created and run. Use a timeout of 300 seconds or more for complex delegation tasks.

​
Next Steps