See at a glance which files an AI coding agent rewrote most during a session by reading the chart in the sidebar.
Compare how many lines were added vs deleted per file using the volume bars on each candle.
Identify unstable files that were heavily edited back and forth by an AI agent across multiple rounds.
README and interface are written in Chinese, non-Chinese users need a translation tool for the setup instructions.
KAgent is a visual extension for the Cursor and VS Code code editors that turns an AI agent's file edits into candlestick stock charts, the kind you'd see on a financial trading platform. Every file in your project becomes a "stock", the first time an AI agent touches that file counts as its "IPO", and each round of edits becomes one candle on the chart. The chart's open and close values represent the line count at the start and end of that edit, while the volume bar shows how many lines were added or deleted. The practical purpose is to give you a visual read on what an AI coding agent actually changed during a session. Instead of reading through a diff, you can glance at the sidebar and see which files had the most turbulent activity, which ones grew steadily, and which ones were heavily rewritten. The extension stores all event data locally in a hidden folder (.kagent/) inside your project, so nothing leaves your machine and the chart panel works offline. Setting it up involves three steps: installing the extension (available from Open VSX or as a downloadable VSIX file), installing the project hooks through a command in the editor's command palette, and then opening the KAgent sidebar panel. Once the hooks are in place, Cursor's afterFileEdit event automatically records each agent edit without any further action needed. The chart panel lets you switch between two color conventions: the Chinese stock market style (red for price rising, green for falling) and the US market style (the reverse). You can click any file in the list to open it in the editor, and the chart view scrolls to follow the latest candle as new edits come in. The README and most interface text are written in Chinese, so non-Chinese readers will need to rely on translation for the setup instructions. The core concept and installation steps are straightforward even so.
← jstone2934 on gitmyhub — every repo by this author, as a profile.
Verify against the repo before relying on details.