Introduction: Why CrewAI + Qwen3?

CrewAI is a Python framework for building structured multi-agent systems that:

• Assign roles like Planner, Researcher, Coder

• Enable collaboration between agents

• Integrate tools like web search, code exec, and APIs

• Work with OpenAI-compatible LLMs—including Qwen3

Paired with Qwen3-Coder, you get a local, powerful, fully open-source multi-agent system ready for real-world automation.

---

1. Setup: What You Need

bash
                           
pip install crewai langchain openai

To use Qwen3 with your local vLLM server:

bash
                           
export OPENAI_API_BASE=http://localhost:8000/v1
export OPENAI_API_KEY=qwen3-key

---

2. Define Agents with CrewAI

Each agent gets a:

• Role

• Goal

• Backstory

• Toolset (optional)

• LLM backend (e.g., Qwen3)

python
                           
from crewai import Agent
from langchain.chat_models import ChatOpenAI

llm = ChatOpenAI(temperature=0.2, model_name="qwen-14b")

planner = Agent(
    role="Planner",
    goal="Break down complex tasks into executable subtasks",
    backstory="An expert strategist who organizes multi-step tasks",
    llm=llm
)

researcher = Agent(
    role="Researcher",
    goal="Find and summarize data from the web",
    backstory="A top analyst who retrieves high-quality, factual information",
    llm=llm
)

---

3. Define Tasks for Each Agent

Tasks are modular and agent-specific:

python
                           
from crewai import Task

plan_task = Task(
    description="Create a plan to build a dashboard using NASA open APIs",
    agent=planner
)

research_task = Task(
    description="Find the best NASA API for space imagery and document it",
    agent=researcher
)

---

4. Assemble the Crew and Execute

python
                           
from crewai import Crew

crew = Crew(
    agents=[planner, researcher],
    tasks=[plan_task, research_task],
    verbose=True
)

result = crew.kickoff()
print(result)

✅ CrewAI automatically routes instructions and coordinates dialogue between Qwen3 agents.

---

5. Add Tools and Memory

CrewAI supports:

• langchain.tools for search, APIs, file access

• memory=True to store prior context

• Custom logic between agents

Example: Tool Integration

python
                           
from langchain.tools import DuckDuckGoSearchRun
search = DuckDuckGoSearchRun()

researcher.tools = [search]

---

6. Example Use Case: AI Research Team

Workflow:
Agents pass data > execute tools > refine output > return results to user.

---

7. Real-World Applications

All powered by self-hosted Qwen3 models using vLLM or Transformers.

---

Conclusion: Qwen3 + CrewAI = Open Multi-Agent Brilliance

With CrewAI and Qwen3, you can build:

• ✅ Autonomous task planners

• ✅ Collaborative researchers and coders

• ✅ Domain-specific agent workflows

• ✅ 100% open-source + offline capable setups

Whether you’re automating internal work, building intelligent assistants, or experimenting with LLM collaboration—Qwen3 + CrewAI delivers structured AI teamwork.

---

Resources

• CrewAI GitHub

• Qwen3-Coder Model

• vLLM Deployment Guide

• LangChain Tools & Agents