Uni-Agent: Train Long-Horizon Agents at Scale

Uni-Agent is a framework for training long-horizon agents.

Get started View on GitHub

• Pluggable

  Connect any existing agent harness to RL.

• Extensible

  Customize agents, tasks, tools, and sandboxes through modular interfaces.

• Scalable

  Run thousands of stateful sessions and turn them into training-ready data (for both SFT and RL).

• Reproducible

  Reproduce results with runnable recipes, complete configs, curves, and benchmarks.