Semanticus: the AI-native semantic-model workbench

Score your Power BI & Fabric semantic models for AI-readiness, fix them with one click or with your own AI assistant, and ship them behind a deploy gate that verifies the evidence. Cross-platform, in VS Code. The free tier covers the full single-edit workflow, with no trial clock, no sign-up and no telemetry.

Why

The next consumer of every model you build is an AI: Copilot, data agents, MCP clients. Models built for humans who already know the business (cryptic names, missing descriptions, no synonyms, undocumented relationships) make AI hallucinate. Semanticus makes "ready for AI" measurable, and gives you and your AI assistant one live workbench to get there.

What you get, free

• Open anything. Power BI Desktop models live, PBIP/TMDL/BIM files offline, or Fabric/Premium XMLA endpoints read-write. Windows, macOS and Linux; the engine is bundled, so there's no .NET install.
• An A to F AI-readiness score. Deterministic, explained findings across naming, descriptions, linguistic schema, Prep-for-AI settings and scale limits, with one-click safe fixes.
• Full model editing. The model tree with a rich Properties grid, a real DAX editor (autocomplete, go-to-definition, formatting), an M editor with interactive applied steps and incremental-refresh setup, an ER diagram with bus-matrix layout, calculation groups, calendars, perspectives, and RLS/OLS roles.
• Lineage, impact and safe-to-remove. Report-aware (PBIR folders or cloud workspaces), down to page, visual and field, with honest tri-state verdicts.
• DAX Lab. Run, benchmark cold/warm, profile, view filter context, and prove rewrites equivalent against live data.
• Best Practice Analyzer. TE-rule-compatible scans, single fixes, honest waivers.
• Docs, diff and deploy. Branded documentation export, model compare and merge, and staged deploys behind a gate that lists exactly what would block.
• DaxLib package manager. Install community DAX UDF packages atomically.

AI-native, by design

Connect your own AI assistant (Claude Code or any MCP client) with one command: Semanticus: Connect Claude Code. It operates the same live session you're looking at. 230+ operations, every change broadcast to the UI as it lands, attributed on one undoable timeline. The Edit History tab shows who did what, you or the assistant, with "Undo to here".

The engine holds no API keys, performs no inference, and sends no telemetry. AI work happens through your assistant on your account.

Pro: enforcement and bulk

Free does everything one step at a time. Semanticus Pro (US$79/yr launch pricing, or US$10/mo) adds the one-click bulk engine and the referee:

• Atomic change plans. Review a batch, apply it as one undoable transaction.
• One-click bulk. Fix-all, apply-all safe fixes, make-AI-ready.
• Enforced workflow runs. 21 shipped playbooks where the engine verifies every gate with evidence: DAX probes, equivalence proofs, re-scans, benchmark deltas. Skipped reads "skipped", never "passed".
• Rule-level waivers and the hash-chained audit-trail export.

Licenses verify offline, so activation works air-gapped.

Quick start

1. Install the extension. Open the Semanticus activity-bar view.
2. Open a model: a running Power BI Desktop instance, a PBIP/TMDL folder, or an XMLA endpoint.
3. Open Studio for the scorecard, diagram, lineage and the rest, or stay in the tree.
4. Optional: run Semanticus: Connect Claude Code to let your AI assistant co-edit the session.

Privacy

Everything runs on your machine. The engine makes no network calls except the data connections you explicitly open (XMLA/Fabric) and optional DaxLib package downloads. No account, no telemetry, ever.

Links: Website · Pricing · GitHub · Support