CrewAI: Build Multi-Agent Teams Where Each Agent Has a Role and Goal

What Is CrewAI?

CrewAI is a Python framework for building multi-agent AI systems where each agent has a defined role, goal, and backstory. Rather than writing a single monolithic prompt, you decompose complex work into a crew of specialized agents that collaborate sequentially or hierarchically.

The mental model is close to a real team: a researcher, a writer, and an editor each do distinct work, hand off results, and produce a final output that none could produce alone.

Core Concepts

Agent — defined by role (job title), goal (what it optimizes for), and backstory (context that shapes its behavior). The backstory is surprisingly powerful for steering tone and expertise level.

Task — a discrete unit of work assigned to an agent, with a description, expected output, and optional context from previous tasks.

Crew — the collection of agents and tasks, plus a process (sequential or hierarchical) that defines execution order.

from crewai import Agent, Task, Crew, Process

researcher = Agent(
    role="Senior Research Analyst",
    goal="Uncover cutting-edge developments in AI and data science",
    backstory="You work at a leading tech think tank. Your expertise lies in "
              "identifying emerging trends and synthesizing complex information.",
    verbose=True,
    allow_delegation=False,
)

writer = Agent(
    role="Tech Content Strategist",
    goal="Craft compelling content on tech advancements",
    backstory="You are a renowned Content Strategist, known for your insightful "
              "and engaging articles. You transform complex concepts into clear prose.",
    verbose=True,
    allow_delegation=True,
)

research_task = Task(
    description="Conduct a comprehensive analysis of the latest advancements in AI. "
                "Identify key trends, breakthrough technologies, and potential industry impacts.",
    expected_output="Full analysis report with trends, technologies, and impacts",
    agent=researcher,
)

write_task = Task(
    description="Using the researcher's findings, develop an engaging blog post "
                "that highlights the most significant AI advancements.",
    expected_output="Full blog post of at least 4 paragraphs",
    agent=writer,
)

crew = Crew(
    agents=[researcher, writer],
    tasks=[research_task, write_task],
    process=Process.sequential,
    verbose=True,
)

result = crew.kickoff()

Memory System

CrewAI supports three memory types: short-term (in-context for the current run), long-term (persisted across runs via SQLite), and entity (knowledge about specific people, places, and concepts extracted from interactions). Long-term memory lets agents build up institutional knowledge over repeated use.

Tool Integration

Agents can use tools including web search (via SerperDev or Tavily), code interpreter (runs Python in a sandbox), file read/write, and custom tools you define as Python functions decorated with @tool.

CrewAI vs AutoGen vs LangGraph

CrewAI is the most approachable for role-based workflows with clear handoffs. AutoGen excels at code generation with back-and-forth agent conversation. LangGraph gives you the most control via explicit state machines but requires more code. For document research + writing pipelines, CrewAI is typically the fastest to build.

Resources

• CrewAI GitHub
• Official docs
• Examples