AutoGen vs CrewAI: which multi-agent framework should you use?

from autogen import ConversableAgent, GroupChat, GroupChatManager
import os

config_list = [{
    "model": "gpt-4o",
    "api_key": os.environ["OPENAI_API_KEY"],
}]

# Define agents with explicit capabilities
researcher = ConversableAgent(
    name="Researcher",
    system_message="You research topics thoroughly and cite sources.",
    llm_config={"config_list": config_list},
)

writer = ConversableAgent(
    name="Writer",
    system_message="You write clear, engaging reports based on research.",
    llm_config={"config_list": config_list},
)

reviewer = ConversableAgent(
    name="Reviewer",
    system_message="You review reports for clarity and accuracy.",
    llm_config={"config_list": config_list},
)

# Group chat for orchestration: AutoGen handles message routing
groupchat = GroupChat(
    agents=[researcher, writer, reviewer],
    messages=[],
    max_round=10,
)

manager = GroupChatManager(groupchat=groupchat, llm_config={"config_list": config_list})

# Explicit message loop: you control the flow
researcher.initiate_chat(
    manager,
    message="Research best practices for Python async/await and provide 3 key points.",
)

from crewai import Agent, Task, Crew
from crewai_tools import tool
import os

# Declarative agent definitions with built-in role support
researcher = Agent(
    role="Researcher",
    goal="Research topics thoroughly and cite sources.",
    backstory="You are an expert research analyst.",
    llm="gpt-4o",
)

writer = Agent(
    role="Writer",
    goal="Write clear, engaging reports based on research.",
    backstory="You are a skilled technical writer.",
    llm="gpt-4o",
)

reviewer = Agent(
    role="Reviewer",
    goal="Review reports for clarity and accuracy.",
    backstory="You are a meticulous editor.",
    llm="gpt-4o",
)

# Declarative tasks: CrewAI chains them automatically
task_research = Task(
    description="Research best practices for Python async/await and provide 3 key points.",
    agent=researcher,
    expected_output="A detailed research report with 3 key points."
)

task_write = Task(
    description="Write a report based on the research findings.",
    agent=writer,
    expected_output="A polished 500-word report."
)

task_review = Task(
    description="Review the report for clarity and accuracy.",
    agent=reviewer,
    expected_output="Reviewed report with feedback."
)

# Crew orchestrates tasks sequentially
crew = Crew(
    agents=[researcher, writer, reviewer],
    tasks=[task_research, task_write, task_review],
    verbose=True,
)

result = crew.kickoff()
print(result)