The Atlan MCP Cookbook

Get build-ready for a new dashboard

Before you build a dashboard: the trusted source, the confirmed definition, and how the data is really used — as the build recipe.

All recipes

On the menu

Use case
Find & understand data
Action
Look up
Persona
Data/BI Analyst, AI Builder, Data Engineer
Effort
Moderate
Runs in
any AI tool

The task

Scenario. You've got a request for a product-sales dashboard. Before opening your BI tool, get build-ready: find the certified table to build on (not a lookalike), confirm the metric definition against what's governed, and read how people already query that table so the dashboard matches real usage.

The ask

  • The trusted, certified source to build on
  • The governed definition — does it match how the team calculates it?
  • The joins, filters, and questions the table already answers

Hero prompt

Hero prompt

Get me build-ready for the product-sales dashboard — the certified source to build on, the Revenue definition, and how people actually query that table.

Step by step

  1. Find the trusted source.

    Step 1

    Find the certified table for <topic> — prefer a mart or gold layer, and steer me off lookalikes.

  2. Confirm the definition.

    Step 2

    Pull the governed term for <metric> on that table and tell me if it matches how we calculate it.

  3. Read how the table is used.

    Step 3

    Show the popular joins, common filters, and the business questions this table already answers.

What you'll give it

Required context

  • The dashboard topic / the metric it's about
  • The team's stated calculation, if you have it, to confirm against the governed one

Optional context

  • Any cut you know you need (by region, by product), so the filters come back relevant
  • Works from a ticket, a Slack ask, or ad hoc

What Atlan creates

Outputs

  • The one certified source to build on — its layer, owner, and certification — with a note to avoid the un-trusted lookalikes
  • The governed definition with a verdict: confirmed — matches, or a flagged difference (tax handling, one-time items, gross vs. net)
  • The build recipe from SQL Intelligence: popular joins (with the key and when to use), common filters, and the business questions the table already answers

Summary. You start building on the certified source with the official number and the questions the data already answers — so the dashboard is right the first time, and a definition mismatch is caught before building, not after.

Read-only

Nothing is written — it stops at build-ready. You build in your BI tool from there.

Follow-up prompts

Keep going once the first answer lands.

Map it to tiles

Map it to tiles

Turn those business questions into a dashboard layout — which become KPI tiles, which a trend, which a detail table.

Get the cut

Get the cut

Which filter do I use for a by-region view, and how is it applied?

Avoid the trap

Avoid the trap

For the top lookalike, show why it's not the trusted one — uncertified, wrong grain, or stale.

Tips & troubleshooting

Cook's note

SQL Intelligence is the hero move — how the table is really queried is the recipe for the dashboard.

  • Match the tool to the work. Any AI tool works; it reads and prepares, so it's safe to run for any request.
  • Automate or chain it. Run it the moment a dashboard request lands, so every build starts from the certified source and the real usage.
  • It only reads. No writes — it stops at build-ready and never builds or changes anything.
THE ATLAN MCP COOKBOOK

Point your AI at the Atlan MCP.
Run your first recipe.

[Website env: production]