[Enhancement] AI Auto-fix flow with Claude Code

[LinceMathew]

Summary

Based on recent feedback, we've been exploring ways to reduce the manual work required to fix issues found by git-lrc.

Currently, the workflow requires users to manually copy and paste each issue into their AI agent's window. This manual step is time-consuming, especially for PRs with a lot of comments.

To tackle this, we are experimenting with an auto-fix method using Claude Code. This allows developers to visually select bugs in our UI and instantly hand them off to Anthropic's Claude Code CLI, which automatically writes the fixes directly into the codebase.

How the New Workflow Works:

• When viewing the git-lrc Web UI, there is now a "Select" option next to every issue.
• The developer can visually review the findings and select only the specific fixes they want to apply.
• Upon clicking "Send to Agent", git-lrc directly sends a JSON payload to Claude Code in the background, and it instantly starts fixing the code.

Here is a demonstration of the flow.

https://www.loom.com/share/6667cccdd31b40079b410aaa74dd2451

Future works

• Make the flow compatible with other AI agents aswell (e.g., Codex, Gemini CLI etc.)
• Option to directly 'send to a JSON file' so that it can be passed to any agents.

We would love to hear your thoughts and feedback on this workflow!

[LinceMathew]

This feature has been implemented and rolled out in this version, v0.3.0.

It is now available in the latest release.

Here is a video showing how it works:
https://www.loom.com/share/b665e3880c3c478499afb9c93d23939f

Please test it out and let us know your thoughts.