🦉 OWL: Optimized Workforce Learning for General Multi-Agent Assistance in Real-World Task Automation

---

中文阅读 | Community | Installation | Examples | Paper | Citation | Contributing | CAMEL-AI

🏆 OWL achieves 58.18 average score on GAIA benchmark and ranks 🏅️ #1 among open-source frameworks! 🏆

🦉 OWL is a cutting-edge framework for multi-agent collaboration that pushes the boundaries of task automation, built on top of the CAMEL-AI Framework.

Our vision is to revolutionize how AI agents collaborate to solve real-world tasks. By leveraging dynamic agent interactions, OWL enables more natural, efficient, and robust task automation across diverse domains.

📋 Table of Contents

• 📋 Table of Contents
• 🔥 News
• 🎬 Demo Video
• 🛠️ Installation
  • Clone the Github repository
  • Set up Environment
  • Install Dependencies
  • Setup Environment Variables
• 🚀 Quick Start
• 🧪 Experiments
• ⏱️ Future Plans
• 📄 License
• 🖊️ Cite
• 🔥 Community
• ⭐ Star History

🔥 News

• [2025.03.07]: We open-source the codebase of 🦉 OWL project.

🎬 Demo Video

371254613005d51d73c82424e56a1d22.mp4

d106cfbff2c7b75978ee9d5631ebeb75.mp4

🛠️ Installation

Clone the Github repository

git clone https://github.com/camel-ai/owl.git
cd owl

Set up Environment

Using Conda (recommended):

conda create -n owl python=3.11
conda activate owl

Using venv (alternative):

python -m venv owl_env
# On Windows
owl_env\Scripts\activate
# On Unix or MacOS
source owl_env/bin/activate

Install Dependencies

python -m pip install -r requirements.txt
playwright install

Setup Environment Variables

In the owl/.env_example file, you will find all the necessary API keys along with the websites where you can register for each service. To use these API services, follow these steps:

1. Copy and Rename: Duplicate the .env_example file and rename the copy to .env.
2. Fill in Your Keys: Open the .env file and insert your API keys in the corresponding fields.
3. For using more other models: please refer to our CAMEL models docs:https://docs.camel-ai.org/key_modules/models.html#supported-model-platforms-in-camel

Note: For optimal performance, we strongly recommend using OpenAI models. Our experiments show that other models may result in significantly lower performance on complex tasks and benchmarks.

🚀 Quick Start

Run the following minimal example:

python owl/run.py

🧪 Experiments

We provided a script to reproduce the results on GAIA. You can check the run_gaia_roleplaying.py file and run the following command:

python run_gaia_roleplaying.py

⏱️ Future Plans

• Write a technical blog post detailing our exploration and insights in multi-agent collaboration in real-world tasks.
• Enhance the toolkit ecosystem with more specialized tools for domain-specific tasks.
• Develop more sophisticated agent interaction patterns and communication protocols

📄 License

The source code is licensed under Apache 2.0.

🖊️ Cite

If you find this repo useful, please cite:

@misc{owl2025,
  title        = {OWL: Optimized Workforce Learning for General Multi-Agent Assistance in Real-World Task Automation},
  author       = {{CAMEL-AI.org}},
  howpublished = {\url{https://github.com/camel-ai/owl}},
  note         = {Accessed: 2025-03-07},
  year         = {2025}
}

🔥 Community

Join us for further discussions!

⭐ Star History