Camel AI

What this article covers

• What Camel AI is and what it’s used for
• How to install Camel AI
• How to build a simple role-based conversation
• How to integrate local models via Ollama
• Who Camel AI is a good fit for

Introduction

Camel AI is a framework for role-based multi-agent conversations. You define two or more agents with distinct roles and let them interact with each other. The classic examples include a buyer negotiating with a seller, a teacher working with a student, or specialists collaborating within a team. Camel AI shines when you want to observe how agents conduct negotiations or exchange knowledge in a given scenario.

Camel AI at a glance

Camel AI operates around roles, tasks, and chat sessions. Each agent receives a system prompt that defines its role. A task sets the topic. The agents converse with each other until a defined endpoint is reached.

Installation

pip install camel-ai

To use it with OpenAI, you’ll need an API key:

export OPENAI_API_KEY="your-key"

For Ollama or other OpenAI-compatible APIs, also set:

export OPENAI_BASE_URL="http://localhost:11434/v1"

Getting started: buyer and seller

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

buyer = ChatAgent(
    system_message=BaseMessage.make_assistant_message(
        role_name="Buyer",
        content="You want to buy a laptop as cheaply as possible."
    )
)

seller = ChatAgent(
    system_message=BaseMessage.make_assistant_message(
        role_name="Seller",
        content="You want to sell a laptop as expensively as possible."
    )
)

message = BaseMessage.make_user_message(
    role_name="Buyer",
    content="What's the price of the laptop?"
)

response = seller.step(message)
print(response.msg.content)

This example starts a conversation between a buyer and seller. You can adapt the roles and topics to suit your needs.

When should you use Camel AI?

Use case	Benefits
Role-playing	Simulate sales, negotiations, coaching
Knowledge transfer	Have agents from different specialties exchange information
Data generation	Generate conversations for training or testing
Research	Study agent behavior in specific scenarios

Common pitfalls

• Roles need to be clear - The more precise your system prompts, the better the conversation flows.
• Local models can struggle - Extended negotiations require a capable model.
• API costs - Many rounds with multiple agents rack up tokens quickly.

Further reading

• CrewAI
• AutoGen
• What is an AI agent?

FAQ - Common questions

Is Camel AI free?

Yes, the open-source framework is free. Costs only apply when you use model APIs.

Can I run Camel AI locally?

Yes, with Ollama or another OpenAI-compatible endpoint. The API key can be a placeholder.

Sources

• Camel AI GitHub
• Camel AI Documentation