whats-new

Metabase 64 and 65? Data apps, Explorations, MCP, Metabot, and much more

· 45 minutes

 

Guests

Andrew Fleming

Andrew Fleming

Product Advocate, Metabase

Abed Habli

Abed Habli

Product Advocate, Metabase

Pierina Salinas

Pierina Salinas

Product Advocate, Metabase

Summary

Two releases in 45 minutes.

Pierina and Andrew walk through what landed in Metabase 64 and 65: AI that digs into your data for you, tighter control over who can use it, a release focused on security, and data apps, custom applications that live inside Metabase.

  • Research mode. Pierina asks a broad question about why sales look flat across categories, and Metabot builds a research plan, runs the charts in parallel, and lets her comment on and collect what’s interesting into a summary. Research needs an AI provider and at least one metric.
  • MCP server, rebuilt. Rebuilt from scratch in 64, the MCP server now goes beyond questions and draft dashboards: AI tools can add tabs, filters, and headings, and create alerts, subscriptions, metrics, and measures. Admins choose which MCP tools each group gets, so a team can read without writing, or write without SQL.
  • Bring your own model. Connect the AI provider you want, self-hosted included, and pick a default model and a smaller one for quick tasks like naming conversations.
  • Security hardening. Two-factor auth is now on Starter too, with an option to require it for everyone on Pro and Enterprise. Cloud also gets IP allowlists and tighter permission checks.
  • Transforms. Indexes now survive rebuilds, and tests catch breaks before they spread.
  • Monitor and the embedding hub. Monitor puts dependency diagnostics, erroring questions, alerts, background jobs, logs, sessions, API keys, and AI usage in one place. The embedding hub does the same for setting up embedding: security, authentication, permissions, appearance, and new embeds.
  • Data apps. Andrew shows data apps, React apps built with AI on top of your semantic layer and managed in Data Studio. His examples range from a 3D asteroid explorer to a live sales tracker to a Kanban board. Data apps can write back to Postgres and MySQL. You build them with the Metabase CLI and its skills, then ship them through Git sync, with no extra servers needed.