---

name: granola-sync description: | Manage Granola meeting transcript extraction and sync. Use when:

• User asks about meeting transcripts, notes, or summaries
• User wants to check sync status or logs
• User needs to re-extract meetings or troubleshoot sync
• User mentions "granola", "meeting notes", or "meeting transcripts" Provides access to extracted meetings in ~/jpt/meetings/ and controls the launchd sync service.

---

Granola Sync

Extracts meeting transcripts and AI-generated summaries from Granola's local cache to markdown files.

File Locations

• Summaries: ~/jpt/meetings/summaries/ (small ~1-2KB, read first)
• Transcripts: ~/jpt/meetings/transcripts/ (large ~30-50KB, read for detail)
• People profiles: ~/jpt/people/ (auto-generated from meetings)
• Extraction script: ~/jpt/meetings/extract_granola.py
• Post-processing prompt: ~/jpt/meetings/post_meeting_prompt.md
• Sync log: ~/jpt/meetings/granola-sync.log
• launchd plist: ~/Library/LaunchAgents/com.jesse.granola-sync.plist
• Granola cache: ~/Library/Application Support/Granola/cache-v3.json

Output Files

Each meeting produces up to two files in separate directories:

• summaries/YYYY-MM-DD_Meeting_Title_summary.md - AI-generated summary with action items (~1-2KB)
• transcripts/YYYY-MM-DD_Meeting_Title.md - Full transcript with metadata and notes (~30-50KB)

For agents: Read summaries first for quick context. Only access transcripts when deeper detail is needed.

Common Tasks

Check sync status

Copylaunchctl list | grep granola-sync
tail -20 ~/jpt/meetings/granola-sync.log

Manual extraction

Copy# Incremental (last 30 hours, skip existing)
python3 ~/jpt/meetings/extract_granola.py

# Full re-extraction
python3 ~/jpt/meetings/extract_granola.py --full

# Specific time window
python3 ~/jpt/meetings/extract_granola.py --hours 48

Service management

Copy# Stop
launchctl unload ~/Library/LaunchAgents/com.jesse.granola-sync.plist

# Start
launchctl load ~/Library/LaunchAgents/com.jesse.granola-sync.plist

List available meetings

Copypython3 ~/jpt/meetings/extract_granola.py --list

# List extracted summaries
ls ~/jpt/meetings/summaries/

# List extracted transcripts
ls ~/jpt/meetings/transcripts/

Script Options

Post-Meeting Processing

When a NEW summary is extracted (first time seen), the script automatically runs Claude to:

1. Update people profiles in ~/jpt/people/ - creates/updates profiles for each attendee
2. Extract action items to Notion Tasks database - adds clear commitments to task list

The prompt is in ~/jpt/meetings/post_meeting_prompt.md - edit it to customize the behavior.

Use --no-process to skip this step (useful for bulk imports or re-extraction).

Automation

The sync runs automatically via launchd:

• On file change: Triggers when Granola's cache updates (after meetings)
• Daily at 6 AM: Backup run in case file watch misses something
• Throttled: Won't run more than once per 60 seconds

Troubleshooting

No new meetings extracted: Check if meetings are within the 30-hour window. Use --hours 48 or --full.

Service not running: Check launchctl list | grep granola-sync. Reload with the service management commands above.

Missing summaries: Granola only caches summaries for meetings you've opened in the app. Open the meeting in Granola first.