Building Multi-Agent Systems with CrewAI

Single agents are powerful, but teams of agents are unstoppable. Here’s how to build multi-agent systems with CrewAI.

Why Multi-Agent Systems?

Just like human teams, AI agent teams can:

• Specialize: Each agent masters one skill
• Collaborate: Share information and delegate tasks
• Scale: Handle complex workflows

Setting Up CrewAI

pip install crewai crewai-tools langchain-openai

Building Your First Crew

Let’s build a content creation team:

Step 1: Define Your Agents

from crewai import Agent
from langchain_openai import ChatOpenAI

llm = ChatOpenAI(model="gpt-4")

# Research Agent
researcher = Agent(
    role='Senior Research Analyst',
    goal='Uncover cutting-edge developments in AI',
    backstory='''You are an expert researcher with a keen eye
    for emerging trends. You excel at finding unique angles
    and backing claims with data.''',
    verbose=True,
    llm=llm,
    tools=[search_tool, web_scraper]
)

# Writer Agent
writer = Agent(
    role='Content Strategist',
    goal='Create engaging content that drives action',
    backstory='''You are a master storyteller who transforms
    complex topics into compelling narratives. Your content
    always includes practical takeaways.''',
    verbose=True,
    llm=llm
)

# Editor Agent
editor = Agent(
    role='Chief Editor',
    goal='Ensure content is polished and error-free',
    backstory='''You have decades of editorial experience.
    You catch every typo, improve clarity, and ensure
    content meets the highest standards.''',
    verbose=True,
    llm=llm
)

Step 2: Define Tasks

from crewai import Task

research_task = Task(
    description='''Research the latest AI agent trends for 2026.
    Focus on: new frameworks, enterprise adoption, and use cases.
    Include statistics and expert opinions.''',
    agent=researcher,
    expected_output='Comprehensive research report with sources'
)

writing_task = Task(
    description='''Write a 1500-word blog post based on the research.
    Make it engaging, include code examples, and add a clear CTA.''',
    agent=writer,
    expected_output='Complete blog post in markdown format'
)

editing_task = Task(
    description='''Review and polish the blog post.
    Fix any errors, improve flow, and ensure SEO best practices.''',
    agent=editor,
    expected_output='Final polished blog post ready for publication'
)

Step 3: Create and Run the Crew

from crewai import Crew, Process

content_crew = Crew(
    agents=[researcher, writer, editor],
    tasks=[research_task, writing_task, editing_task],
    process=Process.sequential,  # Tasks run in order
    verbose=True
)

result = content_crew.kickoff()
print(result)

Advanced Patterns

Hierarchical Process

Let a manager agent delegate work:

crew = Crew(
    agents=[manager, researcher, writer],
    tasks=[complex_task],
    process=Process.hierarchical,
    manager_llm=ChatOpenAI(model="gpt-4")
)

Memory & Context

Enable agents to remember past interactions:

from crewai import Memory

crew = Crew(
    agents=[...],
    tasks=[...],
    memory=True,  # Enable memory
    embedder={
        "provider": "openai",
        "config": {"model": "text-embedding-3-small"}
    }
)

Best Practices

1. Clear Role Definitions - Each agent should have a distinct purpose
2. Specific Backstories - Helps the LLM stay in character
3. Tool Selection - Only give agents the tools they need
4. Task Dependencies - Use context to pass info between tasks
5. Error Handling - Implement retries and fallbacks

Need Help With Your AI Project?

I help businesses build production-ready AI systems. Get in touch to discuss your project!