CAMEL
An open-source multi-agent framework for researching how agent behavior scales, providing agents, tasks, prompts, models and simulated environments for building multi-agent systems.
CAMEL is an open-source framework and research community focused on how multi-agent systems behave at scale. It implements and supports various types of agents, tasks, prompts, models and simulated environments for that research.
Its design centers on four principles: evolvability, where multi-agent systems generate data and interact with environments to improve through reinforcement learning with verifiable rewards or supervised learning; scalability to systems with millions of agents; statefulness, where agents keep memory across multi-step interactions; and code-as-prompt, where code and comments are written to double as instructions for agents.
It supports data generation, task automation and world simulation use cases, is written in Python, and is released under the Apache 2.0 license.