The Short Version
Enzyme uses AI to understand your notes. When you initialize and generate catalysts, note excerpts are processed by our AI providers. When you push to our servers, you can explore your thinking conversationally from anywhere.
Your raw files stay on your device. We store catalysts and embeddings—distilled excerpts, not your full notes. We don't sell your data, we don't train AI on it, and we don't share it.
Our Commitment
You're trusting us with your thinking—notes that contain personal reflections, half-formed ideas, things you're still figuring out. We don't take that lightly.
- ✓ US-based AI providers only. We use AI providers headquartered in the United States with enterprise-grade privacy commitments. Your data never routes through servers in jurisdictions with weaker privacy protections.
- ✓ No training on your data. Our AI providers contractually commit to not using API data for model training. Your notes stay yours.
- ✓ Enzyme as the trust layer. We vet our providers and maintain strict data handling agreements. You don't need to evaluate every AI company—that's our job.
How It Works
Enzyme helps you explore your own thinking, but that requires AI to understand what you've captured. Here's what happens at each step:
When you initialize
You run enzyme init. AI analyzes your vault structure to understand what patterns exist, what entities matter, and how to organize insights. This sends excerpts of your notes to our AI providers. There's no fully offline initialization; AI understanding is what makes Enzyme work.
When you generate catalysts
Enzyme distills your notes into catalysts using AI. Your note content is sent to our providers for processing. Per our agreements, this data is not stored or used for training.
When you push (opt-in)
You run enzyme push to sync catalysts to our servers. This enables conversational exploration of your vault via MCP, so AI can help you multiply your ideas from anywhere.
What We Collect
Account Information (when you sign up)
- Google account email (via OAuth 2.1)
- Authentication tokens (for secure access)
- Subscription status
Pushed Data (only if you run enzyme push)
- Catalysts derived from your notes
- Relevant note excerpts
- Entity references (tags, concepts)
- Semantic embeddings for similarity search
What never gets stored on our servers
- Your full, raw markdown files
- Your full vault structure
Anonymous Analytics
- App version, OS type
- Feature usage patterns (not content)
- Error and crash reports (no personal data)
How We Protect Your Data
Your pushed data is protected by multiple layers:
Encrypted in Transit
All connections use TLS 1.3. Data is encrypted between your device and our servers.
Encrypted at Rest
Our database provider encrypts stored data on disk using AES-256. This protects against physical disk access and backup theft. Encryption keys are stored separately from data.
What at-rest encryption does NOT protect against: application-level access. Database administrators with appropriate access could technically read your data. This is standard for managed database services.
User Isolation
Row-Level Security policies ensure API requests only return your own data. Even with valid credentials, users cannot access each other's catalysts.
Secure Authentication
OAuth 2.1 via Google with automatic token refresh. No passwords stored on our side.
What we don't claim
We don't offer zero-knowledge or end-to-end encryption. To provide AI integration, our servers must process your catalysts. We can technically read what you push. We don't, except to debug issues you report. But we won't pretend otherwise.
AI Processing
Our AI Providers
We exclusively use AI providers that meet our standards:
- Headquartered in the United States
- Enterprise-grade data handling agreements
- Contractual commitment to not train on API data
- SOC 2 certified or equivalent security standards
We regularly evaluate our providers and will update this policy if we make changes. Your data never goes to providers in jurisdictions with weaker privacy protections.
Local Embeddings (optional)
Semantic search can use local models (ONNX/CoreML) that run entirely on your device. This means similarity search works without external API calls.
Remote MCP (opt-in via push)
When you connect your AI assistant to your vault:
- AI reads catalysts you've pushed to our servers
- AI generates responses based on your captured thinking
- This is the whole point: AI that understands your notes
The value exchange: You share your thinking with AI so it can help you explore it. Enzyme doesn't work without AI processing; that's the product.
Your Rights & Controls
Delete Everything
Run enzyme push --clear to remove all pushed data from our servers. This is immediate and permanent. Your local vault is unaffected.
Revoke MCP Access
Disconnect AI from your vault through your account settings. Pushed data remains until you explicitly delete it.
Stay Local
You can use Enzyme indefinitely without ever creating an account or pushing data. Local search, local embeddings, local insights.
Use Your Own AI Provider
If you prefer your data never touch Enzyme's servers, you can use your own AI credentials:
ANTHROPIC_API_KEY=<your-key> OPENAI_API_KEY=<your-key> enzyme init When you provide your own credentials, Enzyme connects directly to your AI provider. No data routes through our servers for initialization or catalyst generation.
GDPR/CCPA
We comply with privacy regulations. Email support@enzyme.garden for data requests.
Contact
Questions about our privacy practices? Reach us at support@enzyme.garden.