Data Vault overview
The Data Vault is where a project’s data lives. It holds the structured tables your agents and apps read and write, the documents you upload, the knowledge stores that power retrieval, and the assets your generated apps reuse.
Every area of the Data Vault is scoped to a single project — data never leaks between projects. You open it from the Data Vault group in the project sidebar.
The four areas
| Area | What it holds | Used by |
|---|---|---|
| Tables | Structured relational tables — the database behind your agents and apps | Database tools, generated apps, deployments |
| Documents | A versioned file store for uploads and agent artifacts | Document tools, RAG sources |
| Content Store | Knowledge stores for retrieval-augmented generation (RAG) | Agent knowledge search, agent memory |
| Assets | Reusable brand and reference files the AI reuses across apps | App generation (Vibe Code) |
How your data reaches agents and apps
The Data Vault stores your data; agents and apps reach it through the platform’s tools and features:
- Tables are read and written with the built-in database tools (
db_query,db_get_schema, and the write tools), and by vibe-coded apps through the app database tier. - Content Stores are searched with the RAG tools and feed an agent’s Episodic and Semantic memory.
- Documents can be linked into a Content Store so their text becomes searchable knowledge.
- Assets are cited in generation prompts with an
@tokenso the AI reuses your logo, brand kit, or reference files.
Shared vs dedicated schema
Tables and Content Stores live in a database schema. Every project starts with a Shared schema. On higher plans you can also create a Dedicated schema for stronger isolation.
- Shared — the default. Your project’s tables are isolated from other projects by row-level security.
- Dedicated — a separate schema for your project, available on higher plans. Where more than one schema exists, a schema selector appears at the top of the Tables and Content Store screens.
Row-level security is always on. Data-access tools run in the calling user’s context and honor any row filters and column masks configured for their role, so agents only ever see the rows they’re allowed to. You can preview this from Manage table data.
Where to start
- New to the Data Vault? Start with Tables — most projects begin by modeling their data.
- Building a knowledge base for an agent? Go to Content Store.
- Reusing brand and reference files across generated apps? See Assets.