CrewAI Setup Guide

Step 1: Install CrewAI

# Create virtual environment
python -m venv venv
source venv/bin/activate

# Install CrewAI (v1.14.5+)
pip install crewai crewai-tools

Step 2: Build a Crew (Updated for v1.14.5)

from crewai import Agent, Task, Crew, Process
from langchain_openai import ChatOpenAI

# Initialize LLM
llm = ChatOpenAI(model="gpt-4o")

# Create agents with roles
researcher = Agent(
    role="Senior Research Analyst",
    goal="Discover insights and trends in AI",
    backstory="You are an expert at analyzing complex data and identifying patterns.",
    verbose=True,
    allow_delegation=False,
    llm=llm
)

writer = Agent(
    role="Content Strategist",
    goal="Create compelling content from insights",
    backstory="You are skilled at turning complex insights into clear narratives.",
    verbose=True,
    allow_delegation=False,
    llm=llm
)

# Create tasks with clear expected outputs
research_task = Task(
    description="Research the latest AI trends in 2026. Focus on orchestration tools and agent frameworks.",
    expected_output="A list of 5 key trends with brief descriptions",
    agent=researcher
)

write_task = Task(
    description="Write a blog post about the research findings.",
    expected_output="A 500-word blog post with introduction, body, and conclusion",
    agent=writer
)

# Create the crew (uses AgentExecutor by default in v1.14.5+)
crew = Crew(
    agents=[researcher, writer],
    tasks=[research_task, write_task],
    process=Process.sequential,
    verbose=True
)

# Run the crew
result = crew.kickoff()

print(result)

Note: In crewAI v1.14.5+, you no longer need to specify CrewAgentExecutor. The crew now uses AgentExecutor by default.

Step 3: Advanced Features

Restore from Previous State

# Resume a crew from a saved state
result = crew.kickoff(restore_from_state_id="your-state-id")

Check Crew Status

# New endpoint format (v1.14.5+)
# GET /status/{kickoff_id}
import requests
response = requests.get(f"https://api.crewai.com/status/{kickoff_id}")
status = response.json()