Query AI grounded in a specific spec page or model object.
Asking AI from the context of a spec or object keeps answers grounded. Instead of describing your system from scratch, you point AI at the relevant page or object and ask your question there.
What grounding provides
When AI is anchored to a spec or object, it can use:
- The object’s metadata, links, and attributes.
- The spec’s prose, decisions, and constraints.
- Attachments (OpenAPI, ERD).
- Related objects in the neighborhood.
Typical questions
| Goal | Example question |
|---|---|
| Understand | ”What does this service depend on?” |
| Extend | ”Add an endpoint to this API and link the data model.” |
| Review | ”Are there orphan objects or cycles near here?” |
| Document | ”Draft an overview spec for this subsystem.” |
Flow
Open the context
Start from the relevant spec page or model object.
Ask a focused question
Reference the object or spec rather than re-describing the system.
Let AI use model tools
AI reads structured context (object, links, docs) to answer accurately.
Review the proposal
For changes, review the diff before accepting.
Structured access
Through the MCP integration, AI reads objects, links, and docs as structured data and can run searches over the model — so answers reflect the real architecture.
If an answer drifts, narrow the context. Ask about one object or one spec at a time for the most reliable results.
Last updated on