The Short Version
Enzyme runs locally on your machine. Your vault files stay on your
device unless you explicitly publish a hosted vault. When you
initialize or refresh catalysts, selected note excerpts are sent to an
LLM provider for analysis. By default, Enzyme uses hosted credits/auth
and ignores inherited personal LLM env keys. If you intentionally pass
--use-env-llm
with your own API key, requests go directly to your provider. Hosted
credit requests are routed to OpenRouter, a US-based model routing provider.
Enzyme does not sell your data or use it to train Enzyme models. Provider processing follows the provider path described below. Hosted features use your workspace, auth, API keys, and credits; local-only artifacts remain on your device.
Data Flow at a Glance
| Action | Stays on device | May leave device | Stored by Enzyme? |
|---|---|---|---|
| Install / agent skill setup | CLI, AGENTS.md/skill files, local config | No note content | No |
| Scan and local embeddings | Files, embeddings, SQLite index | No API call for search/embedding lookup | No |
| Catalyst generation | Generated catalysts and scores after init | Selected excerpts to Enzyme hosted credits/auth by default, or your env LLM provider with --use-env-llm | No raw vault storage |
| Petri / catalyze during a session | Local retrieval over catalysts and source files | No LLM call from Enzyme retrieval | No |
| Hosted workflows / publish | Local artifacts remain removable | Only what you explicitly publish or connect | Yes, for hosted features you enable |
How It Works
Enzyme uses AI to understand your notes. That requires sending
selected excerpts to an LLM provider. Default init/refresh
calls use Enzyme-managed credentials and workspace credits, ignoring inherited
personal LLM env keys. Bring-your-own-key calls go directly to the
provider you configure only when you pass --use-env-llm.
When you initialize
You run enzyme init. AI analyzes your vault structure to understand what patterns
exist and how to organize catalysts. Selected excerpts of your notes
are sent through Enzyme hosted credits/auth by default, or to your
env LLM provider when you intentionally run with
--use-env-llm.
When you generate catalysts
Enzyme distills your notes into catalysts using AI. Selected note excerpts go to the active provider path for processing. Catalysts and embeddings are stored locally on your machine.
When you use the Claude Code plugin
Enzyme ships as a Claude Code plugin. When Claude invokes the /enzyme skill, it calls the Enzyme CLI on your machine to search your local
catalysts. Nothing leaves your device at this step.
What We Collect
Anonymous Analytics
- App version, OS type
- Feature usage patterns (not content)
- Error and crash reports (no personal data)
What never touches our servers
- Your vault files
- Your catalysts and embeddings
- Your API keys
- Raw note files or full vault contents. Selected excerpts may be transmitted through Enzyme hosted credits/auth by default, or to your env LLM provider when you pass
--use-env-llm.
AI Processing
Hosted credits by default — BYOK by flag
When you provide your own API credentials and pass
--use-env-llm,
Enzyme connects directly to your AI provider. We don't proxy,
intercept, or store these calls. Without that flag, default
init/refresh
stay on Enzyme hosted credits/auth even if env keys exist.
OPENAI_API_KEY=<your-key>OPENAI_BASE_URL=<your-endpoint>OPENAI_MODEL=<model-name> enzyme init --use-env-llm
Set OPENAI_API_KEY
or OPENROUTER_API_KEY
to bring your own provider, then pass
--use-env-llm
on enzyme init
or enzyme refresh.
Without that flag, Enzyme uses hosted credits/auth routed through OpenRouter, a US-based model routing provider. Hosted usage may count
against workspace credits.
Local Embeddings
Semantic search uses a compiled-in local embedding model that runs entirely on your device. Similarity search works without any external API calls.
Your Rights & Controls
Local artifacts
Your catalysts, embeddings, and vault index live on your machine.
Delete a vault index by removing /path/to/vault/.enzyme. Delete global Enzyme state, auth, and config by removing ~/.enzyme. Your markdown files are unaffected.
Switch providers anytime
You're not locked into any AI provider. Change your API key and run
init/refresh
with --use-env-llm
to use the new provider. Omit the flag to return to Enzyme hosted
credits/auth. Your local data stays the same.
Workspace features
Current CLI login provisions hosted auth for init credits, publish, and hosted workflows. Local vault artifacts remain removable from your machine.
GDPR/CCPA
Since we don't collect personal data beyond anonymous analytics, there's little to request. But if you have questions, email support@enzyme.garden.
Contact
Questions about our privacy practices? Reach us at support@enzyme.garden.