microsoft / agent-lightning
The absolute trainer to light up AI agents.
Why this job: works in the areas we cover, but does not say which job it does
What it does
- 3,500 lines of code. We treat simplicity as the first principle.
- Train with real agent harnesses. Agents interact with the model through the Agent Lightning v1.0 proxy with ZERO changes, while keeping tools, context, control flow, and environments in the loop.
- Native Kubernetes support. Run agents directly as Kubernetes Jobs without relying on external sandbox services.
- Full coding agent training example. Using only 6K training samples, an end-to-end Qwen3.5-9B workflow improves SWE-bench Verified from 41.8% to 56.4%, a gain of 14.6 percentage points. We release the full pipeline, including data cleaning,
Get started
cd <this-repo>
uv sync
bash scripts/setup_verl.sh 0.8.0 cu130 - Maintenance
- Active
- Last commit today
- Stars
- 18,947
- +43 this week
- Licence
- MIT
- Allows business use
- Downloads for
- None listed
- No release files
- Latest version
- v1.0.3
- Released 9 days ago
- Built with
- Python
- 99% of the code
Is it looked after?
- Contributors
- 24
- Last 90 days
- Issues
- 42% closed
- 117 still open
- Pull requests
- 325 merged
- 82 waiting
- Releases
- 11
- 5 in the last year
- No contributing guide
- Security policy (opens in new tab)
- Code of conduct (opens in new tab)
Similar repos
| Repo | Stars | Last commit | Licence | Downloads for |
|---|---|---|---|---|
| mlop-ai / mlop | 393 | 8 months ago | Apache-2.0 | No downloads |
| meta-pytorch / tnt | 1,723 | 3 days ago | Custom | No downloads |
| flyteorg / flyte | 7,835 | today | Apache-2.0 | No downloads |
More: languages, topics and project record
Languages
- Python 99%
- Shell 1%
Topics
- agent
- agentic-ai
- llm
- mlops
- reinforcement-learning
Project record
- JobOther AI tools
- StatusListed
- Commits687
- Forks1,716
- Watching80
- Stars in 28 days+259
- CreatedJun 18, 2025
- First seen by usNov 7, 2026
- CheckedNov 12, 2026