Home/Agentic AI/CrewAI Basics/Building Crews

CrewAI Basics

Master CrewAI framework for orchestrating role-playing autonomous AI agents

Your Progress

0 / 5 completed
1
2
3
4
5

Putting It All Together: Building Crews

Now that you have agents (roles) and tasks (jobs), you combine them into a Crew. The Crew object orchestrates everything: assigns tasks to agents, manages execution flow, and returns final output.

🚀 Creating a Crew

from crewai import Crew, Process

# Define your crew
crew = Crew(
    agents=[researcher, writer, analyst],  # List of agents
    tasks=[research_task, writing_task, review_task],  # List of tasks
    process=Process.sequential,  # How to execute tasks
    verbose=True  # Show detailed logs
)

# Kick off the crew
result = crew.kickoff()
print(result)  # Final output from last task

Interactive: Crew Execution Simulator

⚙️ Crew Configuration Options

process (Process enum)

How tasks are executed: Process.sequential (one at a time) or Process.hierarchical (manager coordinates).

💡 Start with sequential - it's simpler and uses fewer tokens

verbose (boolean)

Set to True to see detailed logs of agent thinking and tool usage. Great for debugging, but can be noisy.

💡 Use verbose=2 for even more detailed output

manager_llm (LLM object)

For hierarchical process, specify which LLM the manager agent should use. Can be different from worker agents' LLMs.

💡 Example: Use GPT-4 for manager, GPT-3.5 for workers to save costs

memory (boolean)

Enable agents to remember previous interactions across multiple crew runs. Useful for iterative workflows and maintaining context.

💡 Set to True for conversational or multi-session workflows

📦 Complete Example

from crewai import Agent, Task, Crew, Process
from crewai_tools import SerperDevTool

# 1. Define Agents
researcher = Agent(
    role='Senior Researcher',
    goal='Uncover cutting-edge AI developments',
    backstory='You work at a leading tech think tank',
    tools=[SerperDevTool()],
    verbose=True
)

writer = Agent(
    role='Tech Content Strategist',
    goal='Craft compelling content on tech advancements',
    backstory='You are a renowned content creator',
    verbose=True
)

# 2. Define Tasks
research_task = Task(
    description='Research latest AI trends for 2024',
    expected_output='3-paragraph report on AI trends',
    agent=researcher
)

writing_task = Task(
    description='Write engaging blog post about AI trends',
    expected_output='4-paragraph blog post in markdown',
    agent=writer,
    context=[research_task]  # Can access research output
)

# 3. Create Crew
crew = Crew(
    agents=[researcher, writer],
    tasks=[research_task, writing_task],
    process=Process.sequential,
    verbose=True
)

# 4. Run Crew
result = crew.kickoff()
print(result)

🎯 CrewAI Best Practices

  • •Match agents to tasks: Assign tasks to agents with relevant roles and expertise
  • •Start simple: Begin with 2-3 agents and sequential process before adding complexity
  • •Use verbose for debugging: Set verbose=True to understand agent behavior and troubleshoot
  • •Leverage context: Use task context to pass information between agents efficiently
  • •Test iteratively: Run crew with small tasks first to validate agent behavior before scaling
← Prev