Agent Lightning is an MIT-licensed Python package that provides infrastructure for applying reinforcement learning and related optimization methods to AI agents. It is intended for developers and researchers training agentic workflows and integrating them with existing agent systems.
In the Frameworks & runtimes space, Agent Lightning takes a focused approach. It focuses on training and optimizing AI agents with reinforcement learning without building the supporting infrastructure from scratch. Agent Lightning is an open-source project aimed at AI researchers and developers building agentic systems. Agent Lightning is open source under the MIT license. It ships for the command line, and it can be self-hosted.
Microsoft builds and maintains Agent Lightning, and it first shipped in 2025. The project is developed in the open on GitHub with 17.6k stars and 54 commits in the last 90 days. Key capabilities include agent training, reinforcement learning, and agent optimization.
Summary written by a language model from the project’s public pages.
What PulseGate has recorded for this listing
Closest matches by what these projects do