Run saved Metabase questions and ad-hoc MBQL/SQL queries, export results as CSV/JSON/XLSX, and browse collections, dashboards, databases and search.
Metabase ships in the w6w first-party pack. It declares 17 actions, 3 health checks, and the host runs its code in a sandbox that never sees the credential.
io.w6w.metabaseMetabase is a BI tool people point at a database to build questions, save them into collections and pin them onto dashboards — and this app’s job is running a question someone already built and getting today’s number out of it. Run a saved question or a dashboard card with its filters applied, or run raw SQL or MBQL directly against a connected database, and export any result as CSV, JSON or XLSX.
Beyond running numbers, list and search dashboards, collections and databases, read a dashboard’s own cards, and pull a database’s table and field metadata from Metabase’s last sync — useful for building the rest of a workflow around real question and database ids rather than guessing at them. Create new questions and collections directly from a workflow, too.
Metabase is open source and works the same way whether it is self-hosted or running as Metabase Cloud, so this app connects to whichever instance a site actually runs, using an API key scoped to a single permissions group.
Three routes to the same 17 actions. The Workflow tab is generated from Metabase's own manifest and carries its real ids, so it is copy-pasteable; the Code and CLI examples are the same call for any action on any app, so every app-specific value in them is a blank you fill in.
collection-items List the questions, dashboards, models and sub-collections inside a collection. Paginated.
collection-list List the collections this connection can see. Returns a bare array.
dashboard-card-run Run one card as it appears on a dashboard, with the dashboard's filters applied — not the same as running the underlying question.
dashboard-get Fetch one dashboard with its card placements (`dashcards`), tabs and filter definitions.
database-list List the databases registered in Metabase. The `id` of each is what Run Query needs.
database-metadata Fetch one database's tables and fields as Metabase's last sync recorded them — the source for MBQL table and field ids.
query-export Run an ad-hoc native SQL or MBQL query and return the full result set as CSV, JSON or XLSX. Unlike Run Query this is not row-capped.
query-run Run an ad-hoc native SQL or MBQL query against a database registered in Metabase. Truncated at 2,000 rows (10,000 if aggregated) — use Export Query for the full result set.
question-export Run a saved question and return the full result set as CSV, JSON or XLSX. Unlike Run Question this is not row-capped.
question-get Fetch one saved question's full definition — its query, its parameters and its column metadata.
question-list List saved questions. Returns every match with no pagination — use Search for a bounded, filterable list on a large instance.
question-run Run a saved Metabase question and return its rows. Truncated at 2,000 rows (10,000 if aggregated) — use Export Question for the full result set.
question-update Update a saved question's name, description, query, collection or archived state. Only the fields you supply are changed.
search Search questions, dashboards, collections, models, tables and more. The only paginated listing endpoint Metabase offers.
A workflow step names the app and the action, and the editor fills in the
connection when you pick one. This is the Step shape from the
workflow spec, carrying Metabase's real ids.
{
"manifestVersion": "2",
"name": "metabase-example",
"steps": [
{
"id": "collection-create",
"uses": {
"app": "io.w6w.metabase",
"action": "collection-create",
"connection": "conn_YOUR_CONNECTION_ID"
},
"with": {
"name": "<name>"
}
}
]
}collection-create collection-items collection-list dashboard-card-run dashboard-get +12 more actions available
Every app-specific value here is a blank you have to fill in. An
app action is reached through the connection that authenticates it, so the
address is a connection id, not the app id — and connections belong to your account,
so a public page cannot know yours. Create one for Metabase, then fill in
the three blanks: conn_YOUR_CONNECTION_ID, the action key, and the
parameters that action declares. The call itself is real — the shape is transcribed
from the studio's own snippet builder, which prints the same kind of blanks — but
nothing in it is specific to Metabase. The Workflow tab is where this app's
real ids are.
npm install @w6w/sdkyarn add @w6w/sdkpnpm add @w6w/sdkdeno add npm:@w6w/sdkimport { W6wClient, isActionRun } from "@w6w/sdk";
// Reads W6W_BASE_URL and W6W_TOKEN from the environment when omitted.
const client = new W6wClient();
const envelope = await client.run({
urn: "conn_YOUR_CONNECTION_ID",
action: "query-run",
payload: {
database: "<value>",
type: "<value>",
query: "<value>",
// parameters: "<value>",
},
});
if (isActionRun(envelope)) console.log(envelope.value); npm install -g @w6w/cli w6w run conn_YOUR_CONNECTION_ID --action query-run --payload '{"database":"<value>","type":"<value>","query":"<value>"}' Give an AI agent Metabase — without giving it Metabase's credentials. One MCP endpoint exposes every app, function and workflow the caller is entitled to, as tools it can discover and run. Access is granted per team while we onboard.
One tool call{
"name": "w6w_invoke",
"arguments": {
"ref": "app:io.w6w.metabase#collection-create",
"input": {
"name": "<name>"
}
}
}
Every tool names its target with a single ref. The
app: form above doesn't name a connection at all — the
host resolves which of the caller's Metabase connections to sign
with, and refuses rather than guesses when the answer is ambiguous.
The token is attached host-side, at the moment of the call. It is never a tool argument, never in the model's context, and never in a transcript — so a prompt injection has nothing to exfiltrate.
Tools are derived per end user from what that person has actually connected and is entitled to — not one shared bot identity carrying the union of everyone's access.
Multi-step work runs on the workflow engine and returns a run handle the agent can poll — retries, branching and state survive the conversation that started them.
Metabase's declared health checks are on the surface too, so an agent can tell "the vendor is down" from "your credential expired" before it burns a retry on either.
The MCP surface is part of the hosted platform. Metabase itself is MIT, and the runtime that executes it is source-available (FSL).
Metabase declares its own checks, so its health is a property of the app rather than something the host guesses at.
Component status from status.metabase.com (Atlassian Statuspage). Covers Metabase's own hosting and store — a self-hosted instance is unaffected, which is why this check is informational and the per-connection `instance` check carries the weight.
Unauthenticated `GET /api/health` against this connection's Metabase — proves the host resolves, Metabase is serving, and its application database answers. Deliberately not `/livez`, which returns 200 unconditionally and cannot report a broken instance.