microsoft/agent-lightning
The absolute trainer to light up AI agents.
About the project
A lightweight (~3,500 LOC) reinforcement-learning framework for training AI agents with their real harnesses: agents run unchanged while a proxy gateway captures training data. Includes Kubernetes support and coding-agent training examples.
Useful for
- Train a coding agent on your own repositories and tests via RL
- Run agent rollouts as Kubernetes Jobs without external sandbox services
- Collect a trajectory dataset from agent model calls via the API Gateway
README summarized by DeepSeek V4.1 Flash. Details may be inaccurate.
Why it’s trending
- 20 stars so far today, about 26 expected by the end of the day.
- Over the last two days the pace is 3.4× that of the previous week and a half.
- GitHub Trending Python today: #13, +70 stars.
- About 8 forks a day — people are taking the code.
Stars per day
Bars are daily stars, the line is the usual pace. Red marks spike days.
Numbers
- Total stars
- 18,336
- Stars in a day
- 26
- Forks
- 1,615
- Issues and pull requests
- 591
- Watchers
- 84
- Language
- Python
- License
- MIT
- Latest release
- v1.0.1 · August 24, 2026
- Created
- June 18, 2025
- Last push
- September 17, 2026
Star trust
Growth looks organic: forks and discussion are in line with active projects, and stars arrive unevenly, the way people give them.
These are heuristics, not a verdict: we judge by the repository’s behavior, not by a list of stargazers.
Spotted in
- September 18, 2026GitHub Trending Python today: #13, +70 stars
Similar projects
-
8.8
browser-use/jev-ultrafast
A browser agent with a dynamic, indexed action space: given one natural-language goal it picks an operation (CLICK, TYPE_TEXT, SELECT, etc.) and an element, using a small LLM to generate text only for typing.
-
8.5
tamaratran/fast-jev-compaction
Claude Code plugin that replaces context compaction summaries with Jev decisions: every tool call and result is scored in one fast request, stale ones are dropped or truncated, everything kept stays verbatim. Also…
-
7.9
hypit-ai/hypit
Hypit gives AI agents a language and system to create video: the agent clones a viral clip as an editable workflow — footage, captions, B-roll and effects anchored to words instead of seconds.
-
7.9
Tencent/BrowserSkill
BrowserSkill connects AI agents (Cursor, Claude Code, Codex, etc.) to your already logged-in browser via the bsk CLI and an extension, letting the agent work in a separate window without interrupting you.
-
7.0
korcarc/text-humanizer
A Python tool that rewrites AI-generated text through an LLM and translation pipeline to bypass detectors like Turnitin and GPTZero. Supports 8 languages and requires a DeepSeek API key.
-
6.8
Tencent/WeKnora
Open-source LLM knowledge platform: turns raw documents into queryable RAG, an autonomous reasoning agent, and a self-maintaining Wiki.