CrewAI is a Python framework for orchestrating role-based AI agents that collaborate on tasks. It is one of the hottest multi-agent frameworks in 2026 alongside LangGraph. If you see ModuleNotFoundError: No module named ‘crewai’, install with one pip command.
Step 1: Install crewai
# Base install:
pip install crewai
# With tool integrations (RAG, file I/O, web search):
pip install 'crewai[tools]'
# Specific Python version (CrewAI needs 3.10+, 3.13 not yet stable):
python3.11 -m pip install crewai
Step 2: Minimal crew example
from crewai import Agent, Task, Crew, Process
researcher = Agent(
role='Research Analyst',
goal='Find recent AI breakthroughs',
backstory='Expert at scanning technical papers and summarizing key findings.',
verbose=True
)
writer = Agent(
role='Tech Writer',
goal='Turn research into clear blog posts',
backstory='Award-winning explainer of complex tech.',
verbose=True
)
research_task = Task(
description='Find 3 recent advances in multimodal LLMs',
agent=researcher,
expected_output='3 bullet points with sources'
)
write_task = Task(
description='Write a 500-word blog from the research',
agent=writer,
expected_output='Final blog post in markdown'
)
crew = Crew(
agents=[researcher, writer],
tasks=[research_task, write_task],
process=Process.sequential,
verbose=True
)
result = crew.kickoff()
print(result)
Step 3: Use a local Ollama model (no API cost)
from crewai import Agent, LLM
local_llm = LLM(
model="ollama/llama3.1",
base_url="http://localhost:11434"
)
agent = Agent(role='Analyst', goal='...', backstory='...', llm=local_llm)
Why this error happens
| Cause | Fix |
|---|---|
| Not installed | pip install crewai |
| Python under 3.10 | CrewAI requires 3.10-3.12; install correct Python |
| Pinned ancient version | pip install -U crewai |
| No tools, need integrations | pip install ‘crewai[tools]’ |
Frequently Asked Questions
CrewAI vs LangGraph: which should I use?
CrewAI uses a high-level role-based mental model (give each agent a role, goal, backstory). LangGraph is lower-level with explicit state graphs and edges. CrewAI is faster to prototype; LangGraph gives you more control for production deployments with checkpoints and time-travel debugging.
Does CrewAI need OpenAI?
No. CrewAI uses LiteLLM under the hood and supports OpenAI, Anthropic, Gemini, Ollama, Groq, Mistral, Azure, Bedrock, Vertex, and 100+ providers. Set OPENAI_API_KEY or pass an LLM object directly to each Agent.
Is CrewAI free and open-source?
The framework is free (MIT). CrewAI Enterprise (managed hosting + observability + SLA) is a paid product. The library itself works for any project, including commercial.
Can agents use custom tools?
Yes. Define a tool with @tool decorator or BaseTool subclass, then pass it in tools=[] to an Agent. Built-in tools cover file I/O, web search (Serper), CSV/PDF read, RAG over directories, code execution.
What processes does CrewAI support?
Process.sequential runs tasks in order. Process.hierarchical adds a manager agent that delegates to crew members and reviews their output. Both can run with verbose=True for full traces.