Getting Started with Camel-AI: Build Your First Multi-Agent System

Introduction

Camel-AI makes it remarkably simple to create multi-agent systems. In this tutorial, you will build a two-agent system where a Task Planner and a Code Executor collaborate to write and test a Python script — completely autonomously. You will see firsthand how role-playing agents divide work, communicate, and deliver results.

Prerequisites

Before starting, ensure you have:

• Python 3.9+ installed
• An OpenAI API key (or Anthropic/Gemini)
• Basic familiarity with Python

Step 1: Install Camel-AI

pip install camel-ai

This installs the core library along with all dependencies for agent communication, memory, and tool use.

Step 2: Set Your API Key

export OPENAI_API_KEY="sk-..."

Step 3: Create Your First Multi-Agent System

Create a file called multi_agent_demo.py:

from camel.agents import ChatAgent
from camel.messages import BaseMessage
from camel.types import RoleType

# Create two agents with distinct roles
planner = ChatAgent(
    system_message=BaseMessage(
        role_name="Task Planner",
        role_type=RoleType.ASSISTANT,
        content="You are a meticulous task planner. Break down coding tasks into clear, sequential steps.",
    ),
)

coder = ChatAgent(
    system_message=BaseMessage(
        role_name="Python Coder",
        role_type=RoleType.ASSISTANT,
        content="You are an expert Python developer. Write clean, well-documented code.",
    ),
)

# Planner assigns a task
task = "Write a Python function that calculates Fibonacci numbers using dynamic programming."
planner_msg = BaseMessage(
    role_name="User",
    role_type=RoleType.USER,
    content=task,
)
response = planner.step(planner_msg)
print(f"Planner: {response.msg.content}")

# Coder executes
coder_msg = BaseMessage(
    role_name="User",
    role_type=RoleType.USER,
    content=f"Implement this plan:\n{response.msg.content}",
)
code_response = coder.step(coder_msg)
print(f"Coder: {code_response.msg.content}")

Step 4: Run It

python multi_agent_demo.py

You will see the Planner break down the task into steps, and the Coder produce the actual implementation. This simple pattern scales to dozens of agents with specialized roles.

Exploring Further

Camel-AI supports much more than two-agent chats:

• Role-Playing Societies: Create any number of agents with custom personas
• Task-Oriented Loops: Agents can iterate on tasks with feedback loops
• Data Generation: Use built-in pipelines for synthetic data creation
• Tool Integration: Give agents access to external APIs and tools

Conclusion

In just a few lines of Python, you built a functioning multi-agent system. Camel-AI abstracts away the complexity of inter-agent communication, letting you focus on what matters: designing agent behaviors that solve real problems. Check out the official documentation for advanced topics like memory, tool use, and multi-turn conversations.