CodeLens

AI code reviews of your pull requests, from a model on your own PC

CodeLens turns a repository into a code graph, reviews each pull request with a local model and the code around the change, and posts its findings as comments, usually within a minute. Every finding comes with the evidence it was based on. Your code goes between GitHub and your machine, and nowhere else.

Download CodeLens 0.1.1 for WindowsSee the live demoSelf-host (any OS)

106 MB · Windows 10 or 11, 64-bit · Release notes. Self-hosting needs Docker, Ollama and a terminal, on Windows, macOS or Linux.

24 seconds: one changed line, reviewed by a local model, caught, and traced back to the code it cites.

What's new in 0.1.1

Every release
  • Reviews no longer fail when Ollama starts after CodeLens, as it often does when Windows starts: a review waits for it, up to 10 minutes, and says so in its timeline.
  • A review that failed because Ollama wasn't running is reviewed again by itself once Ollama answers.
  • A failed review's page has Review again.
  • Reviews that were waiting when Windows restarted are no longer lost: CodeLens picks them up when it starts again.
  • Uninstalling no longer leaves an empty CodeLens folder behind.
  • A user's guide, from the download to troubleshooting: docs/desktop.md.

Already installed? CodeLens offers each new release itself: Settings → Updates, or the tray.

After the download

  1. Run the installer
    Windows says “Windows protected your PC”, since CodeLens isn't code-signed yet: choose More info, then Run anyway.
  2. Follow the setup
    It checks your PC, installs Docker Desktop and Ollama if they're missing, and downloads the AI models (about 6 GB).
  3. Create your GitHub App
    One click on GitHub makes an App that is yours alone; install it on the repositories you want reviewed.
  4. Open a pull request
    Each push is reviewed, usually within a minute on an NVIDIA GPU, with the findings as comments on the pull request.
What your PC needs
The setup checks all of this before it downloads anything.
  • Windows 10 (22H2) or 11, 64-bit, with virtualization on
  • 8 GB of memory at least; 16 GB is comfortable
  • 15 GB free on the disk at least; 30 GB is comfortable
  • An NVIDIA GPU with 6 GB or more is best; without one, reviews take a few minutes each
Check the download
The file should be exactly the one GitHub built for this release.

In PowerShell, Get-FileHash .\CodeLens-Setup-0.1.1.exe should print:

29C5380554CA5777F34E47C115CFB857637AA3AB1BC82CA9BC40962FB0E4D34A

To remove it later: Settings → Apps → CodeLens → Uninstall. It asks whether to keep your data.