Platform Concepts
These are the product terms you will see in Dataira.
How an answer is made
Key terms
| Term | Meaning |
|---|---|
| Project | A workspace area that groups related datasources and conversations. |
| Datasource | One governed database or bounded API source. |
| Schema review | The step where you confirm what tables and columns mean. |
| Live datasource | A datasource that is ready for chat questions. |
| SQL preview | The generated SQL shown so technical users can inspect the answer. |
| MCP | A way to use Dataira from coding tools like Claude Code or Codex. |
| Analytical Pack | A versioned context that pins metrics and operations to exact source revisions. |
Source scope and end-user access
A Project owns database connections. An Environment owns its OpenAPI connector configuration, API keys, and one published Analytical Pack context. The pack records immutable source bindings; it does not grant access by itself.
At runtime, Dataira intersects the environment-scoped token catalog with the end user's datasource allowlist and governed data policy. A pack, skill, browser, or user prompt cannot widen that result. Per-user restrictions are supported today. Inherited organization defaults are planned; until then, bind each user or role explicitly through the data-policy API.
Remote vs Tunnel
Tunnel is usually best for local development, private networks, and databases behind a firewall.