Using the IDE extension
Screens, reading provenance and risk colour, and day-to-day workflows.
The visual surface. Same engine as the CLI, same numbers, arranged for reading.
First open
Prism indexes the workspace. Until the index exists, screens that need it say so rather than showing zeroes — an empty chart and an unbuilt index look identical otherwise and mean different things.
Screens
| Screen | For |
|---|---|
| Overview | Health, DNA, activity, most-connected files |
| Map | Structure laid out spatially, with layers |
| Domains | Frontend / backend / data / infrastructure when DNA detects them |
| Impact | Blast radius, safe delete, rename impact |
| Trends | Health over time |
| Integrations | Optional connectors, each consented |
| Settings | Indexing, appearance, privacy |
Reading the interface
- Provenance markers — estimated vs unavailable; see signal provenance.
- Risk colour — green → amber → red maps to risk bands.
- Confidence — on features and stack signals; low means look before you act.
Working in it
- Impact → enter a path → blast radius, tests, features.
- Change review reads the working tree for aggregate risk.
- Map zoom: repository → package → feature → file → symbol; layers overlay coupling, churn, tests, risk.
- Right-click: Blast Radius, Safe Delete, Explain, Reveal on Map.
Cursor: extension + agent
Keep the same project open. Add MCP, then ask structural questions in chat while using the map visually. If the agent guesses instead of calling tools, say once: "Use Prism for that."