Camel

CAMEL is an open source, modular framework for building multi agent AI systems in Python. It supplies the primitives for creating agents that reason, plan and call tools, composing them into societies with defined roles, and running them against code interpreters, memory stores and retrieval pipelines. The project describes itself as a community researching the scaling laws of agents for data generation, world simulation and task automation. It suits researchers and engineers who want building blocks rather than a hosted product. Agents are atomic reasoning units driven by large language models; societies are coordinator layers that assign roles, delegate tasks and manage collaboration; interpreters provide Python, shell and browser execution; and memory and storage layers keep chat history, tool outputs and learned knowledge across long tasks. Retrieval augmented generation pipelines combine chunking, retrieval and generation, and synthetic data engines run self instruct, chain of thought and Source2Synth pipelines with verifier loops. Documentation includes a quickstart, an API reference and cookbooks covering data generation, retrieval and simulation. The framework is free and open source, installed from PyPI with an optional bundle of all toolkits and interpreters. The documentation publishes no pricing and no paid tier. Several related projects sit in the same ecosystem: OASIS, a large scale social simulation environment modelled on platforms such as Reddit and Twitter; the CRAB benchmark for cross environment automation tasks on Ubuntu and Android; Project Loong for verifier driven synthetic question and answer data; and OWL, a multi agent automation framework built on top of CAMEL for real world tasks using browsers, code interpreters and multimodal models.