AgentSynth

AgentSynth: Scalable Task Generation for Generalist Computer-Use Agents

Below are instructions to run our pipeline:

Data Generation in OSWorld

Please refer to https://github.com/xlang-ai/OSWorld for environment setup.

To install requirements:

pip install -r requirements.txt

To generate tasks and trajectories:

python generate_and_save_traces_persona.py

InSTA

First clone the repository into the insta folder using the following command:

cd insta_data
git clone https://github.com/data-for-agents/insta.git
cd insta
git checkout $(git rev-list --max-count=1 --before="2025-05-12" main)
cd ..

Then, if you have not run InSTA before, pull their Docker image:

docker pull brandontrabucco/insta-browser-environment

Load your secret API keys in a file called insta_data\secrets.json. Finally, run:

python combined_task_generation_new.py --env insta

For troubleshooting your InSTA installation, refer to their repository at https://github.com/data-for-agents/insta/tree/main.

📄 Citation

If you use AgentSynth in your research, please cite our paper:

@article{xie2025agentsynth,
  title={AgentSynth: Scalable Task Generation for Generalist Computer-Use Agents},
  author={Xie, Jingxu and Xu, Dylan and Zhao, Xuandong and Song, Dawn},
  journal={arXiv preprint arXiv:2506.14205},
  year={2025}
}