CrewAI

Learning Objectives

• Understand what CrewAI is and what problem multi-agent frameworks solve
• Identify CrewAI's core abstractions: agents, tasks, crews, and tools
• Compare CrewAI vs. AG2 and LangChain for multi-agent orchestration

What Is CrewAI?

CrewAI is an open-source multi-agent AI framework created by João Moura in 2024, which quickly became one of the most popular frameworks for orchestrating teams of AI agents. CrewAI's central metaphor is a "crew" — a team of specialized AI agents who each have a defined role, goal, backstory, and toolkit, and who collaborate to complete complex tasks that exceed what a single agent could accomplish.

CrewAI was designed to be simpler and more developer-friendly than competing frameworks like AG2 — focusing on role clarity, readable configuration, and fast time-to-working-agent.

Core Concepts

Agents — Specialized Roles

Each agent in CrewAI is a specialized entity defined by:

• Role: What the agent is (e.g., "Senior Research Analyst", "Content Writer", "QA Engineer")
• Goal: What this agent is trying to accomplish
• Backstory: Additional context that shapes the agent's behavior and personality
• Tools: What capabilities the agent has (web search, code execution, file reading)
• LLM: Which model powers this agent (different agents can use different models)

from crewai import Agent
from crewai_tools import TavilySearchTool

researcher = Agent(
    role="Senior Research Analyst",
    goal="Uncover cutting-edge developments in AI",
    backstory="You work at a leading tech think tank. You're skilled at finding, analyzing, and synthesizing complex information.",
    tools=[TavilySearchTool()],
    llm="gpt-4o",
    verbose=True
)

Tasks — Specific Deliverables

Tasks define specific work to be done, assigned to a specific agent:

from crewai import Task

research_task = Task(
    description="Research the current state of AI agents in enterprise software. Focus on adoption rates, key vendors, and emerging patterns.",
    expected_output="A detailed analysis report with 5–7 key findings, cited with sources",
    agent=researcher
)

Crews — Orchestrated Teams

A Crew assembles agents and tasks into a collaborative workflow:

from crewai import Crew, Process

crew = Crew(
    agents=[researcher, writer],
    tasks=[research_task, writing_task],
    process=Process.sequential  # or Process.hierarchical
)

result = crew.kickoff()
print(result.raw)

Two process types:

• Sequential: Tasks execute in order; each agent builds on the previous task's output
• Hierarchical: A manager agent plans and delegates tasks; agents report back

CrewAI Tools

CrewAI comes with built-in tools and integrates with LangChain tools:

• Search tools: TavilySearchTool, SerperDevTool (Google search)
• File tools: FileReadTool, FileWriteTool, DirectoryReadTool
• Web tools: WebsiteSearchTool, ScrapeWebsiteTool
• Code tools: CodeInterpreterTool
• Database tools: PGSearchTool (PostgreSQL search)
• Custom tools: Define any Python function as a tool with a decorator

Memory and Learning

CrewAI supports different memory types:

• Short-term memory: Context across tasks in a crew run
• Long-term memory: Persistent memory across runs (SQLite-backed)
• Entity memory: Track and recall facts about specific entities
• User memory: Personalization based on past interactions

CrewAI Flows — Event-Driven Orchestration

CrewAI Flows is the production architecture for building and deploying multi-agent systems. Flows support conditional logic, loops, real-time state management, and integration with external systems — letting you start with rules-based steps, add LLM enrichment, layer in agent delegation, and scale to full crew orchestration within the same framework.

The Enterprise AMP Suite adds tracing and observability, a unified control plane for managing AI agents, and seamless integrations with existing enterprise systems.

Comparison: Multi-Agent Frameworks

Pricing

CrewAI is free and open source (Apache 2.0). Costs:

• LLM API calls: charged by the providers you use
• CrewAI+ cloud platform: optional managed hosting and monitoring; pricing available on request

Strengths

• Readable configuration: Agent definitions read like natural language — easy for teams to understand what each agent does
• Role clarity: The role/goal/backstory structure produces more focused, appropriate agent behavior than generic agent definitions
• Fast to prototype: A functioning two-agent crew can be built in under 30 lines of code
• Massive scale: Over 12 million executions per day, roughly 2 billion total agentic executions, used by 60%+ of Fortune 500
• Hierarchical process: Manager agent pattern enables sophisticated task planning without manual orchestration logic
• LLM flexibility: Different agents can use different models — use GPT-4o for reasoning, Haiku for fast summarization

Limitations & Considerations

• Newer framework: CrewAI has grown rapidly but LangChain's ecosystem is still broader
• LLM cost at scale: Multi-agent workflows make many LLM calls — each task is one or more API calls per agent
• Hallucination compounding: Errors in early agent outputs can propagate and compound in later agent tasks
• Coordination overhead: Agents passing context through tasks can introduce latency vs. a single well-prompted LLM call
• Limited streaming support: Real-time streaming of crew output is less developed

Best Use Cases

When to choose alternatives:

• Single-agent with tool use → LangChain ReAct agent
• Complex stateful agent graphs → LangGraph
• Conversational multi-agent (Microsoft ecosystem) → AG2
• Production automation with observability → n8n or LangSmith

Getting Started

pip install crewai crewai-tools

from crewai import Agent, Task, Crew

researcher = Agent(role="Researcher", goal="Find key facts", backstory="Expert analyst")
writer = Agent(role="Writer", goal="Write a clear summary", backstory="Technical writer")

task1 = Task(description="Research AI agent trends in 2026", agent=researcher,
             expected_output="5 key findings with sources")
task2 = Task(description="Write a 300-word summary based on the research", agent=writer,
             expected_output="Clear, engaging summary paragraph")

crew = Crew(agents=[researcher, writer], tasks=[task1, task2])
result = crew.kickoff()
print(result.raw)

Key Takeaways

• CrewAI is a multi-agent framework that defines teams of specialized AI agents with roles, goals, backstories, and tools that collaborate to complete complex tasks
• Sequential and hierarchical process modes allow both ordered task pipelines and manager-delegated agent workflows
• Simpler and more readable than AG2; less powerful for complex stateful graphs than LangGraph
• Free and open source; costs are the underlying LLM API calls only
• Best for role-based task delegation workflows — research/writing pipelines, software development crews, and multi-perspective analysis tasks