Skip to content
| Marketplace
Sign in
Visual Studio Code>Visualization>Codefinds — Typing & AI Code TrackerNew to Visual Studio Code? Get it now.
Codefinds — Typing & AI Code Tracker

Codefinds — Typing & AI Code Tracker

Preview

kabirmirza

|
6 installs
| (0) | Free
See exactly how your code was written. Tracks typed, pasted, and AI-generated code with live line-by-line highlights, stats dashboard, and session reports. 100% local — no data leaves your machine.
Installation
Launch VS Code Quick Open (Ctrl+P), paste the following command, and press enter.
Copied to clipboard
More Info

Codefinds — Typing & AI Code Tracker

See exactly how your code was written — typed, pasted, or AI-generated.


Features

  • Live gutter highlights — Green = typed, amber = pasted, purple = AI-assisted, gray = bulk edit
  • Status bar — Real-time percentage breakdown at a glance: ⚡ 62% typed · 18% pasted · 20% AI
  • Sidebar dashboard — Donut chart, timeline, per-file breakdown, and live activity log
  • Classroom mode — Teachers can monitor student coding activity in real time via a local server
  • WPM tracking — Rolling words-per-minute counter
  • Session persistence — Survives restarts; auto-prunes data older than your retention setting
  • Export reports — Dump session data as Markdown or JSON

How It Works

Codefinds listens to VS Code's document change events and classifies each edit using a multi-signal heuristic:

Category Signal
Typed Single-character insertions at human typing speed (30–500 ms intervals)
Pasted Multi-character single-event insertion that matches clipboard content
AI-assisted Large insertion (>20 chars) not matching clipboard, no active selection, AI extension detected
Bulk edit Multiple ranges changed in one event (multi-cursor, format-on-save, find-replace)

Important: AI detection is heuristic-based, not definitive. It infers intent from patterns — it cannot read Copilot's mind. See Known Limitations.


Classroom Mode

Codefinds includes an optional classroom feature for educators:

  • Students connect to a teacher's local server using a short classroom code
  • Teachers see a live dashboard of all student activity — typing rate, paste %, AI %, flagged students
  • Only metadata is transmitted (character counts and classifications) — actual code is never sent
  • Each student gets a unique identity per workspace and per classroom, so sessions stay isolated

To use classroom mode:

  1. The teacher runs the included codepulse-server Node.js server (node server.js)
  2. Students use Codefinds: Connect to Classroom from the Command Palette and enter the server URL, their name, and the classroom code
  3. The teacher opens http://localhost:3000 to view the live dashboard

Configuration

Setting Default Description
codefinds.enabled true Master on/off switch
codefinds.showGutterDecorations true Show color-coded line highlights
codefinds.showStatusBar true Show stats in status bar
codefinds.aiDetectionSensitivity "medium" low / medium / high — how aggressively to flag AI
codefinds.excludePatterns ["*.min.js", ...] Glob patterns for files to skip
codefinds.sessionRetentionDays 7 Days of history to keep
codefinds.classroomServer "localhost:3000" Address of the classroom server
codefinds.studentName "" Your display name in classroom sessions
codefinds.classroomFlagThreshold 40 AI % above which students are flagged on the teacher dashboard

Commands

Command Description
Codefinds: Toggle Line Highlights Show/hide gutter decorations
Codefinds: Show Dashboard Open the sidebar dashboard
Codefinds: Reset Session Data Clear all tracked data for this workspace
Codefinds: Export Report Export session as Markdown or JSON
Codefinds: Pause/Resume Temporarily pause tracking
Codefinds: Connect to Classroom Join a teacher's classroom session
Codefinds: Disconnect from Classroom Leave the current classroom session

Access all commands via the Command Palette (Cmd+Shift+P / Ctrl+Shift+P).


Privacy

In standalone mode, all data stays on your machine — stored locally in VS Code's workspace storage, never transmitted anywhere.

In classroom mode, only metadata is sent to your teacher's server: character counts and edit classifications (typed / pasted / AI). Your actual code is never transmitted.

No telemetry, no analytics, no crash reports — in either mode.


Known Limitations

  • AI detection is heuristic only. It cannot reliably distinguish between a snippet expansion, a multi-character autocomplete, a large paste that doesn't match the current clipboard, and an actual AI suggestion. It makes an educated guess based on size, timing, and whether AI extensions are installed.
  • Clipboard check is asynchronous. There's a small window where a paste might briefly appear as "AI-assisted" before the clipboard is read and the classification is updated.
  • Undo/redo sequences are skipped and do not re-classify lines.
  • Formatter on save is detected heuristically and classified as bulk_edit. May occasionally misfire on legitimate multi-cursor edits.
  • Snippet expansions (including Emmet) may be classified as ai_assisted if they exceed the AI threshold size.
  • Auto-closing brackets and quotes are classified as typed with slightly lower confidence.

License

MIT — see LICENSE

  • Contact us
  • Jobs
  • Privacy
  • Manage cookies
  • Terms of use
  • Trademarks
© 2026 Microsoft