How AI interacts with your data
OmegaVault is retrieval, not inference. We never train on your memories. Every AI read is explicit, per-call, and revocable.
Retrieval, not inference
OmegaVault does not run its own model. When an AI client asks your vault for context, OmegaVault returns matching memories over MCP. The AI does the reasoning; we provide the memory.
No training on your data
We do not train our models, and we do not license your memories to third parties for training. Your content is the input to your intelligence, not a dataset for ours or anyone else's.
Explicit, per-call reads
A connected client does not get a blanket pull of your vault. Each query asks for specific, relevant context. You can see what's been requested in Connected apps.
Revocable
Revoke a client at any time from Connected apps; it instantly loses access to your vault.
What inference providers see
When your AI client (e.g. Claude) retrieves context, that context is sent to the inference provider you chose (e.g. Anthropic) as part of your normal use of that tool. OmegaVault does not control what happens to context once a third-party model has it. We recommend reviewing that provider's data terms.