Integration

Metabase Automation

There are four questions called “Revenue”, and the one feeding the exec dashboard is the one nobody remembers writing.

A BI tool is only as trustworthy as its housekeeping. AI agents search your Metabase collections, tell you which cards actually feed a dashboard, and build the question you asked for from the real table schema — with your approval before any card or dashboard changes.

agentorg.run / metabase
Metabase logo
Y

You

Message Hattie
How it works

Ask in plain language, get the real numbers

Your agent queries Metabase itself rather than guessing, and shows you what it ran.

Step 1

Connect Metabase

One click, read access first. Your agent can see the same data you can and nothing you can't.

Step 2

Ask the question you actually have

No query language and no dashboard hunting. The agent works out which Metabase data answers it and goes and gets it.

Step 3

Check the working, then share it

You get the answer along with what it queried, so you can sanity-check the number before it goes in front of anyone.

Actions

What your agent can pull from Metabase

  • Run a dataset query and return the result
  • Create, copy, and update questions from the live table schema
  • Assemble a dashboard out of existing cards
  • Search collections to work out which card feeds what
  • Report the databases, tables, and columns available to query
Recurring work

Reporting that stops landing on you

The reports someone rebuilds every week, assembled from the source each time.

Your agent searches every collection for cards with the same name or the same query shape, and tells you which of them a dashboard actually depends on.

Your agent reads the table schema first, writes the query against real columns rather than assumed ones, and saves it as a draft card for review.

Name the metrics that matter to them. Your agent finds the cards that already exist, writes the missing ones, and lays them out in a new dashboard.

Example questions

Three things worth asking on Monday morning

Find the duplicate questions

Your agent searches every collection for cards with the same name or the same query shape, and tells you which of them a dashboard actually depends on.

Build a question from a plain-English ask

Your agent reads the table schema first, writes the query against real columns rather than assumed ones, and saves it as a draft card for review.

Assemble a dashboard for a new team

Name the metrics that matter to them. Your agent finds the cards that already exist, writes the missing ones, and lays them out in a new dashboard.

Who it's for

Teams without an analyst on call

Teams whose analyst left

The dashboards still work but nobody can safely change them. Your agent can read the collections, explain what a card queries, and write the next one — so the instance keeps evolving rather than being frozen.

Ops leads answering the same question weekly

Someone asks for a cut of the numbers that no existing card covers. Instead of a Slack request into a queue, your agent writes the question against the real schema and saves it where the team can find it again.

FAQ

Metabase questions, answered

What people ask before putting an AI agent to work — answered straight.

They can, and the schema-reading is what makes it work. Your agent lists the tables and columns before it writes anything, so the query references fields that exist. The card is saved as a draft for you to run and check before it goes on a dashboard.

Ready to automate Metabase?

Connect your tools, put an agent to work, and approve what matters. Free to start.