Skip to main content

​
Overview

VibeLearn works with Cursor via hook scripts. When a Cursor agent session ends, the same 5-step analysis pipeline runs: stack detection → static analysis → concept extraction → quiz generation → sync.

​
Quick Install

# Clone and build
git clone https://github.com/anergcorp/vibelearn.git
cd vibelearn && bun install && bun run build

# Install hooks for all projects (recommended)
bun run cursor:install -- user

# Start the worker
bun run worker:start
Then restart Cursor to load the hooks.

​
Configure an AI Provider

~/.vibelearn/settings.json
{
  "VIBELEARN_GEMINI_API_KEY": "AIza..."
}
See Gemini Setup or OpenRouter Setup for full guides.

​
What Gets Captured

Cursor HookVibeLearn Action
beforeSubmitPromptInitialize session
afterFileEditCapture file edits
afterMCPExecutionCapture MCP tool usage
afterShellExecutionCapture shell commands
stopTrigger analysis pipeline

​
After a Session

vl status     # Check what was analyzed
vl quiz       # Quiz yourself
vl gaps       # See weak areas

​
How It Differs from Claude Code

FeatureClaude CodeCursor
HooksNative plugin hooksBash scripts in .cursor/hooks/
Transcript accessAvailableNot available
Analysis pipelineFull (with transcript context)Full (file edits only)
Worker port3777837778
Installation/plugin install vibelearnbun run cursor:install -- user
The main limitation: Cursor doesn’t provide transcript access, so concept extraction uses file edit observations only. This is still effective for most sessions.

​
Troubleshooting

Hooks not running: 
chmod +x ~/.cursor/hooks/*.sh
# Check Cursor Settings → Hooks tab
Worker not responding: 
curl http://127.0.0.1:37778/api/readiness
tail -f ~/.vibelearn/logs/vibelearn-$(date +%Y-%m-%d).log
No quiz questions: Verify AI provider key is set in ~/.vibelearn/settings.json and check worker logs for LLM errors.