Your intelligent assistant, built as a Claude Code plugin. Geoffrey learns your preferences and patterns over time, providing personalized assistance across work, travel, and personal tasks.
Named after Geoffrey Hinton (godfather of AI) and Geoffrey from Fresh Prince of Bel-Air.
Version: 0.1.0 Status: In Development
- ✅ Plugin structure
- ✅ Knowledge management skill
- ✅ /preferences command
- ✅ iCloud knowledge storage
- ⏳ Local installation and testing
- 🔜 Travel planning skill (Phase 2)
- 🔜 Team management skill (Phase 3)
- 🔜 Automatic learning from conversations (Phase 4)
- Learning-enabled: Remembers your preferences with confidence scoring
- Skills-based: Capabilities activate automatically based on context
- Privacy-first: Your data in iCloud, code in GitHub
- Multi-machine: Same Geoffrey, same preferences, different devices
- Open source: MIT licensed, community contributions welcome
- macOS (for iCloud sync)
- Claude Code 2.0+
- iCloud Drive enabled
- Git
# 1. Clone repository
git clone https://github.com/krishagel/geoffrey.git
cd geoffrey
# 2. Set up iCloud knowledge directory
mkdir -p ~/Library/Mobile\ Documents/com~apple~CloudDocs/Geoffrey/knowledge
# 3. Add as local marketplace (for development)
/plugin marketplace add ~/non-ic-code/geoffrey
# 4. Install plugin
/plugin install geoffrey@geoffrey
# 5. Verify installation
/preferencesJust tell Geoffrey your preferences in natural language:
> "I always prefer Marriott hotels. I'm Platinum Elite."
Geoffrey: "I've learned your hotel preferences:
- Primary chain: Marriott
- Loyalty tier: Platinum Elite
- Confidence: 1.0 (explicitly stated)
Stored in knowledge base."
/preferences # View all preferences
/preferences travel # View travel preferences only
> "Actually, I prefer aisle seats on flights"
Geoffrey: "Updated your airline preferences:
- Seat preference: Aisle (was: Window)
- Confidence: 1.0"
Geoffrey Plugin
├── Skills (auto-activate)
│ └── knowledge-manager ← Phase 1
├── Commands (user-invoked)
│ └── /preferences ← Phase 1
├── Agents (specialized workers)
│ └── (coming in Phase 2+)
└── Hooks (automation)
└── (coming in Phase 4)
Knowledge Storage (iCloud)
└── ~/Library/.../Geoffrey/knowledge/
├── preferences.json ← Phase 1
├── memory.jsonl ← Phase 4
└── patterns.json ← Phase 4
# 1. Make changes to Geoffrey files
cd ~/non-ic-code/geoffrey
# Edit skills, commands, etc.
# 2. Reload plugin
/plugin reload geoffrey
# 3. Test changes
> "Test the feature"geoffrey/
├── .claude-plugin/
│ ├── plugin.json # Plugin metadata
│ └── marketplace.json # Marketplace config
├── skills/
│ └── knowledge-manager/ # Knowledge management skill
│ └── SKILL.md
├── commands/
│ └── preferences.md # /preferences command
├── agents/ # (Future)
├── hooks/
│ └── hooks.json # Hook definitions
├── .gitignore
└── README.md # This file
preferences.json:
{
"version": "1.0",
"last_updated": "2025-11-17T10:30:00Z",
"preferences": {
"travel": {
"hotels": {
"primary_chain": "Marriott",
"loyalty_tier": "Platinum Elite",
"confidence": 1.0,
"learned_from": ["explicit:2025-11-17"],
"last_updated": "2025-11-17T10:30:00Z"
}
}
}
}- Basic plugin structure
- Knowledge management skill
- /preferences command
- iCloud storage
- Travel planning skill
- Trip planning based on preferences
- OmniFocus integration
- Points optimization
- Team management skill
- Freshservice integration
- Employee support workflows
- Post-conversation hooks
- Automatic preference extraction
- Pattern detection
- Confidence scoring refinement
Geoffrey's skills activate automatically based on what you say:
> "Plan my trip to Seattle"
→ travel-planning skill activates (Phase 2)
> "How is my team doing?"
→ team-management skill activates (Phase 3)
> "I prefer Marriott hotels"
→ knowledge-manager skill activates (Phase 1)
Geoffrey tracks how confident it is about each preference:
- 1.0 = You explicitly told Geoffrey
- 0.8-0.9 = Strong pattern (5+ observations)
- 0.6-0.7 = Moderate pattern (3-4 observations)
- 0.4-0.5 = Weak pattern (1-2 observations)
- <0.4 = Insufficient data
Higher confidence preferences take priority in suggestions.
- Local storage: All data in your iCloud Drive
- No tracking: Geoffrey doesn't send data anywhere
- You own it: Edit or delete knowledge files anytime
- Open source: Audit the code yourself
- Encrypted: Consider encrypting sensitive data (loyalty numbers)
Contributions welcome! Please:
- Fork the repository
- Create a feature branch
- Make your changes
- Test thoroughly
- Submit a pull request
See CONTRIBUTING.md for details (coming soon).
Geoffrey is inspired by:
- Personal AI Infrastructure by Daniel Miessler
- Claude Code's plugin system
- Personal assistant paradigm
MIT License - see LICENSE for details.
- Issues: GitHub Issues
- Discussions: GitHub Discussions
- Email: [email protected]
- Anthropic for Claude Code
- Daniel Miessler for Personal AI Infrastructure inspiration
- The open source community
Built with ❤️ using Claude Code
Current status: Phase 1 implementation Last updated: November 17, 2025