AI agents for Metabase
A Monmo agent works in Metabase like the teammate who knows SQL and your schema. Ask a question in plain words and it builds the query, runs it, checks the result and shows you the SQL. It repairs questions that broke after a schema change, documents models and columns, and writes the commentary your dashboards lack. Saving or changing anything in Metabase asks first.
Updated 2026-11-13 · Free to start. Paid plans from $15 a month.
What agents do in Metabase
Answer a question in plain words
Ask how many customers upgraded last month and it writes the native query or builds it with the query builder, runs it, sanity-checks the totals and shows you exactly what it ran.
AskHow many workspaces upgraded to a paid plan in October, split by signup source? Show me the SQL.
Speed up a slow native query
Give it the question that takes a minute to load. It reads the SQL, finds the join or subquery doing the damage, proposes a faster version and compares the results row for row.
AskThe Weekly Revenue question takes 90 seconds. Rewrite the SQL so it is faster and prove the numbers match.
Repair questions after a schema change
When a column is renamed or a table moves, saved questions start failing. The agent reads each error, finds the change behind it and prepares the fixed queries for you to approve in one batch.
AskFind every question in the Finance collection that errors since the orders table changed, and prepare fixes.
Document models and columns
It reads your tables and models, writes a plain description for each column and metric, and flags fields whose meaning is unclear. Updating the descriptions in Metabase waits for your approval.
AskWrite descriptions for every column in the Subscriptions model and list the ones you could not work out.
Add the story to a dashboard
Every Monday at 9 it reads the KPI dashboard, compares the week with the four before it and drafts a short note for Slack naming the card behind each figure.
AskEvery Monday at 9, read the KPI dashboard and draft a five-line note for #metrics. Ask before posting.
Tidy the collections
It finds duplicate questions, dashboards nobody has opened in months and personal copies of shared reports, then proposes what to archive and where the survivors should live.
AskAudit Our analytics and propose which questions and dashboards to archive or merge.
Set up a parameterised question
It turns a one-off query into a saved question with variables for date range and plan, so a dashboard filter can drive it, and tests each filter value before asking to save.
AskTurn my churn query into a saved question with date and plan filters, then add it to the Retention dashboard.
How it connects
- 1You connect your own Metabase once, on Metabase Cloud or self-hosted, with your sign-in or an API key an administrator creates. A self-hosted instance has to be reachable from the internet for the agent's cloud computer to use it.
- 2You choose which agents may use Metabase. A data or engineering agent belongs there; an agent that only handles email or scheduling does not.
- 3Running read queries, exploring tables and drafting notes need no approval. Saving or editing a question, dashboard, model or column description, and changing alerts or subscriptions, asks first.
More in Analytics and data
Other apps of this kind that agents work in, each with its own page.
Questions people ask
Can an AI agent write SQL in Metabase for me?
Yes. Describe what you want to know and a Monmo AI agent writes the native query against the database you connected to Metabase, runs it and checks the result for obvious mistakes such as double counting. It always shows you the SQL, so you can read it, and saves the question only when you allow it.
Will the agent's queries slow down our production database?
Its queries go through Metabase on the same database connection your team already uses, so they cost what any other question costs. You can ask it to keep row limits and narrow date ranges while it explores. Pointing Metabase at a read replica keeps all exploratory work away from your production database.
Does it work with self-hosted Metabase as well as Metabase Cloud?
It works with both, provided the agent can reach the instance. A Metabase Cloud site connects directly. A self-hosted Metabase on a private network needs a public address or another route your engineers choose. The agent works through the Metabase connection, or in its own browser when the connection does not cover something.
Can it build Metabase dashboards on its own?
It builds everything up to the save. It drafts the questions, picks the visualisations, lays out the cards and wires up the filters, then shows you a card listing what it will create in which collection. Nothing appears in Metabase until you allow it, and you can decline any single card.