Build AI agents with CrewAI

Prerequisites

Step-by-Step

1. 1

   Install CrewAI

   The crewai package ships the core orchestrator. Add crewai-tools for built-in integrations.

   pip install crewai crewai-tools

2. 2

   Define agents

   Each agent is a role + goal + backstory. Think of these as system prompts with structure.

   from crewai import Agent
   researcher = Agent(
     role='Senior Researcher',
     goal='Find the most recent verified facts about {topic}',
     backstory='You are obsessive about citations.',
     llm='gpt-4o-mini'
   )

3. 3

   Define tasks

   Tasks are units of work assigned to agents. Be explicit about expected output format.

   from crewai import Task
   research_task = Task(
     description='Research {topic} and produce a markdown brief with citations.',
     expected_output='A markdown document, 500-800 words, with at least 5 citations.',
     agent=researcher
   )

4. 4

   Assemble the crew

   Sequential process is simplest. Hierarchical lets a manager agent route work dynamically.

   from crewai import Crew, Process
   crew = Crew(
     agents=[researcher, writer],
     tasks=[research_task, write_task],
     process=Process.sequential
   )
   result = crew.kickoff(inputs={ 'topic': 'pgvector vs pinecone' })

5. 5

   Add tools

   Wrap any callable as a tool. Use crewai-tools for search, scraping, file ops out of the box.

   from crewai_tools import SerperDevTool
   researcher.tools = [SerperDevTool()]

6. 6

   Inspect outputs

   kickoff() returns a CrewOutput with raw text, JSON, and per-task results. Log the whole thing during dev.

Common Pitfalls

What's Next

• ->Move to Process.hierarchical for dynamic routing.
• ->Wrap your crew in a FastAPI endpoint for production.

More Build AI agents