Cairn Documentation

The whole model as tables: every entity of every kind, searchable (including the names behind references), sortable, scoped to a branch of the tree, and editable in place. References are links in both directions, and an Integrity view lists every broken reference, loop, or duplicate. Every edit is an ordinary ChangeSet.

Automated checks across nodes, requirements, interfaces, state machines and trace links. Scores the share of the model with nothing to fix, and each finding leads to its source.

A Python notebook on a snapshot of your model. The agent writes and runs cells, reads the figures, and saves the notebook, its conversation, and its runs to the project as you work, so you can reopen them without rerunning. You can keep as many notebooks as you want. Save outputs keeps a file or figure the notebook produced. When the model changes, a saved analysis is marked stale; Update to latest project keeps one earlier result for that notebook and reruns. You bring your own key — the Anthropic key from Settings, or an OpenAI or xAI key entered in the notebook.

Full ChangeSet history with operation-level detail. Every AI edit and manual change is recorded with timestamps, specialist identity, and undo capability. A project started from the interview has an Origin entry: the answers, each actor and requirement the interview settled and what it became, what the first build added and proposed, the device the interview started on, and how the first proposal was settled.

AI cost tracking ledger. Shows total estimated spend, breakdowns by provider/category/model, recent API calls with token counts, and CSV export. Updated automatically on every API call.

Trace link explorer showing satisfies, verifies, derives, and depends_on relationships between requirements, verifications, and nodes.

Unit system management with SI base units, derived units, and custom type definitions for interface signals and properties.

Asset browser for generated files — images, meshes, scripts, and saved notebooks (the notebook, its conversation, and its runs). A notebook card shows whether it matches the current model or is an earlier, stale result. Stored locally in IndexedDB via Dexie.js.

API key configuration (Anthropic, Gemini), model preferences, project metadata, naming conventions, and export options.