data-metabase
Installation
SKILL.md
Metabase Automation
The classic Metabase REST API owns cards, dashboards, collections, permissions, and schema refresh. The versioned Agent API is the surface for headless semantic BI assistants and app-side AI workflows. The official MCP server (added in v60) and @metabase/cli (added in v62) serve AI clients and terminal scripting. Which release added what is in references/metabase-59-surface.md; before citing version-specific behavior, read the user's version from GET /api/session/properties (version), the current stable line from metabase.com/releases, and behavior from /docs/latest.
Quick Reference
| Task | Path | Use When |
|---|---|---|
| Create/update questions and dashboards | Classic REST API + scripts/metabase_api.py |
Standard content automation and incremental upserts |
| Promote content between environments | Remote Sync or serialization | Git-backed promotion, reviewable diffs, cross-environment moves |
| Build embedded customer analytics | Embedding + tenants + embedding permissions | Multi-tenant apps, customer portals, row-level isolation |
| Build an AI analytics app | Agent API | Versioned, semantic, app-side AI querying |
| Integrate Metabase with an AI coding agent | MCP server (v60+) | Claude, Cursor, VS Code — generate questions and dashboards via conversation |
| Govern AI access by group | Metabot AI governance (v61+, Pro/Enterprise) | Per-group controls, token limits, usage analytics |
| Refresh schema metadata | Database sync/rescan endpoints | New tables, changed columns, stale field values |
| Tune native SQL questions | Export-first + native query patterns | Stable automation without guessing request shapes |