How Vault AI works

Models, connectors, your vault, entities and training — the mechanics behind the assistant.

Which AI models can I use in Vault AI?

Claude, Gemini, ChatGPT, Llama and DeepSeek — all from one chat thread. You can switch between them mid-conversation, so you always use whichever model fits the question best instead of juggling five separate apps.

What does a connector (Gmail, Drive, Calendar, Contacts) actually let Vault AI see?

Every connector is opt-in and read-only — off until you turn it on, and scoped only to what that feature needs. Vault AI never posts, edits, or deletes anything in a connected account, and you can disconnect any connector in one tap.

What is the Vault, and how is it different from just chatting?

Your vault is personal storage for files, notes and documents you choose to upload. Once something's in your vault, your AI can answer questions grounded in that content — so replies use your actual context instead of generic knowledge.

What are entities, brains, and skill modules?

An entity is an AI you build from three parts: a brain (which model powers it), a personality archetype (Generator, Adaptive, Search, or Smart Mix), and skill modules (See, Think, Predict, Remember). Mix and match to shape how it responds to you.

What is the Neural Forge / training gauge?

The Neural Forge is where your assistant adapts to your own patterns and data over time — you can watch convergence, epochs and cluster utilization move as it learns. This adaptation is private to your account and is never used to improve anyone else's AI.

Still have a product question?

Ask us directly — we'll reply by email.

Or email us directly at sohalyash85@gmail.com