Looker

You can use the Looker integration to query Looker models and view directly from your Stack AI workflows.

---

Connect to Looker

Prerequisites

• A running Looker instance (Looker Core, Looker (Google Cloud core), or self-hosted)

• Admin access, or permission to manage API credentials for your user account

• Your Looker instance must be network-accessible from Stack AI (not restricted to a private VPN)

Step 1: Generate API Credentials in Looker

Looker API access uses a per-user Client ID and Client Secret, not a shared OAuth app. You generate these in the Looker Admin panel.

If you are a Looker Admin

1. In Looker, go to Admin > Users.

2. Click on the user you want to use for the API connection (create a dedicated service user if preferred).

3. Scroll to the API Keys section and click Edit Keys.

4. Click New API3 Key.

5. Copy both the Client ID and Client Secret. Note: the secret is only shown once.

If you are generating keys for your own account

1. Go to your Looker account settings (click your avatar > Account).

2. Scroll to the API Keys section.

3. Click New API3 Key and copy the Client ID and Client Secret.

For production use, create a dedicated Looker service account user rather than tying the API key to a personal account. If that user is deactivated, the connection will break.

Step 2: Find Your Looker Base URL

Your Base URL is the root address of your Looker instance, without any path suffix.

Looker Deployment Type	Base URL Format	Example
Looker (Google Cloud core)	https://<instance>.looker.com	https://mycompany.looker.com
Looker hosted on GCP	https://<instance>.cloud.looker.com	https://mycompany.cloud.looker.com
Self-hosted	Your custom domain	https://looker.mycompany.com

Stack AI appends /api/4.0 to this URL for all API calls. Do not include a trailing slash or path.

Step 3: Connect in Stack AI

In the Stack AI connector setup, provide the following three fields:

Field	Description
Base URL	Root URL of your Looker instance
Client ID	API key Client ID from your Looker user settings
Client Secret	API key Client Secret (treated as a password)

Click Connect. A green status indicator confirms the connection is healthy.

Looker User Permissions Required

The Looker user whose API key you use must have sufficient permissions to run queries. Looker's permission model is role-based.

Operation	Required Looker Permission
Authenticate (connect)	Any valid API key — no special role needed
Health check (/api/4.0/user)	Any active user account
Create and run queries	access_data + explore on the relevant models
View query results	see_looks or direct model access via model sets

The minimum role for querying data is one that includes access_data and explore on the models you want to query. Your Looker admin configures this through Admin > Roles.

Troubleshooting

401 Unauthorized on connect

• Double-check that the Client ID and Client Secret are copied correctly with no extra whitespace.

• Confirm the user account is active and not locked out.

• Verify the Base URL is correct and reachable. Try opening {base_url}/api/4.0/login in a browser.

403 Forbidden when querying

• The connected user does not have access_data or explore permission on the model/view you are querying. Ask your Looker admin to assign the appropriate role.

Connection times out

• Looker may be behind a firewall or VPN that blocks inbound requests from Stack AI. Ensure Stack AI's IP range is allowlisted in your network configuration.

Token expired mid-workflow

• Stack AI automatically refreshes the token 60 seconds before expiry. If you see auth errors mid-run, it is likely a network issue preventing the refresh call from completing.

---

Available Actions

1. Query Looker

Run a query against a Looker model and view, and return the results as structured data. This is the primary way to pull data out of Looker into a Stack AI workflow.

Input Parameters

Parameter	Required	Type	Description
Model Name	Yes	String	The name of the Looker model to query (e.g. ecommerce, thelook)
View Name	Yes	String	The name of the view within the model (e.g. orders, users)
Fields	Yes	String array	List of fields to return, in view.field_name format (e.g. ["orders.count", "orders.created_date"])
Filters	No	Key-value pairs	Field-level filters to apply (e.g. {"orders.status": "complete", "orders.created_date": "7 days"})
Limit	No	Number	Maximum number of rows to return

Finding model and view names: In Looker, open the Explore menu. The top-level menu item is the model name; the section within the explore is the view name. Field names follow the pattern view_name.field_name.

Output Parameters

Parameter	Type	Description
Query Results	Array of objects	Each object is one row of results, with field names as keys
Field Metadata	Array of objects	Metadata for each field in the result: name, label, and type (dimension, measure, or table calculation)
Row Count	Number	Total number of rows returned

Example

To get the count of completed orders grouped by creation date for the last 7 days:

• Model Name: ecommerce

• View Name: orders

• Fields: ["orders.count", "orders.created_date"]

• Filters: {"orders.status": "complete", "orders.created_date": "7 days"}

• Limit: 500