Looker
Looker: run Looks and inline queries against the modelled data, and inspect dashboards, models, users and schedules.
About
Looker ships in the w6w first-party pack. It declares 11 actions, 2 health checks, and the host runs its code in a sandbox that never sees the credential.
- App id
io.w6w.looker- Version
- 0.1.1
- Author
- w6w
- Licence
- MIT
- Categories
- Analytics & BI · Data Warehousing & ETL
Overview
Looker turns a data warehouse into governed, reusable queries, and this app runs those queries and inspects what’s built around them without touching the warehouse’s raw SQL. Run a saved Look or an ad-hoc query against a model and Explore, and read back the models, dashboards, users and scheduled plans that make up a Looker instance.
Every query respects Looker’s own model — field names stay qualified to their view, and a row limit is required so a workflow can’t accidentally trigger an unbounded warehouse scan. Because access in Looker is scoped per user, the app also surfaces who a credential is and what it can see, which explains why the same query can return different rows through different connections.
This app is read-only by design: dashboards, Looks and scheduled plans are content an analyst owns, so nothing here creates or edits them. Good for pulling metrics into a workflow on a schedule, auditing who has access to what, or checking which warehouse connections a query actually depends on.
Build with Looker
Three routes to the same 11 actions. The Workflow tab is generated from Looker'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.
List database connections
connection-list The warehouses Looker queries — where the cost of every Look actually lands. Reports the dialect, the connection-pool ceiling that queries share with the interface, and whether a query could trigger a derived-table build.
List dashboards
dashboard-list Dashboards visible to this credential's user. Each TILE is its own warehouse query, so a dashboard refresh is many queries — a workflow that wants one number should run a Look.
Describe an Explore
explore-get Every field an Explore exposes, split into DIMENSIONS and MEASURES — selecting only measures returns one row and selecting a dimension groups by it, which is the commonest surprise in a Looker query. Hidden fields are listed, because the API can still select them.
Get a Look
look-get A Look's definition without running it — the model, Explore, fields and filters a workflow is really depending on. Note a SOFT-DELETED Look still answers here, so fetching one is not a test of whether it exists.
List Looks
look-list Saved Looks visible to this credential's Looker USER — two connections legitimately see different sets. Counts the PUBLIC ones, whose URLs need no login, and excludes the soft-deleted ones the API still returns.
Run a Look
look-run Run a saved Look — a saved QUERY, not a saved result, so this hits the warehouse now. Its definition lives in Looker and can change without the workflow changing, so the Look's `updated_at` comes back with the rows.
Get the current user
me-get Who this credential is in Looker. There is no scope on a Looker API key — access is the USER's, including row-level access filters, so the same query run by two credentials can legitimately return different data.
List LookML models
model-list The LookML models and their EXPLORES — the vocabulary `query-run` needs, and neither name is guessable. The Explore names here are what the query API calls `view`; the LookML views underneath are not.
Run a query
query-run Run an ad-hoc query against a model and Explore. It compiles to SQL and runs on the WAREHOUSE, so the cost is somebody else's budget — a positive `limit` is required and Looker's unlimited `-1` is refused.
List scheduled plans
scheduled-plan-list What Looker sends on its own, and where. Each plan is a recurring WAREHOUSE QUERY and a recurring data export — and the destination list is the only place a schedule still going to somebody who left is visible.
List users
user-list Looker users, separating real accounts from EMBED users — which can vastly outnumber them and are not the same thing. Counts who holds API credentials, which is the only record of which integrations exist.
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 Looker's real ids.
{
"manifestVersion": "2",
"name": "looker-example",
"steps": [
{
"id": "explore-get",
"uses": {
"app": "io.w6w.looker",
"action": "explore-get",
"connection": "conn_YOUR_CONNECTION_ID"
},
"with": {
"model": "<model>",
"explore": "<explore>"
}
}
]
}Here are some of the things you can do
-
Describe an Explore
readexplore-get -
List database connections
readconnection-list -
List dashboards
searchdashboard-list -
Get a Look
readlook-get -
List Looks
searchlook-list
+6 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 Looker, 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 Looker. 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: "explore-get",
payload: {
model: "<value>",
explore: "<value>",
// includeHidden: "<value>",
},
});
if (isActionRun(envelope)) console.log(envelope.value);npm install -g @w6w/cliw6w run conn_YOUR_CONNECTION_ID --action explore-get --payload '{"model":"<value>","explore":"<value>"}'Give an AI agent Looker — without giving it Looker'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.looker#explore-get",
"input": {
"model": "<model>",
"explore": "<explore>"
}
}
}
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 Looker connections to sign
with, and refuses rather than guesses when the answer is ambiguous.
What the agent gets
Credentials it can't read
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.
A tool surface scoped to the caller
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.
A durable workflow in one call
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.
Health-aware discovery
Looker'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. Looker itself is MIT, and the runtime that executes it is source-available (FSL).
Health checks
Looker declares its own checks, so its health is a property of the app rather than something the host guesses at.
Looker status
Declared unavailable — there is no Looker service to have a status. Every deployment is its own instance, hosted or self-hosted, so an app-scoped check has nothing to read. The `instance` check answers this for a particular connection.
Looker instance reachable
Reaches this connection's own Looker. Signed, because Looker offers no unauthenticated health endpoint — so it names which failure a given error looks like, including the SELF-HOSTED port 19999, which presents as the instance being down.