> ## Documentation Index
> Fetch the complete documentation index at: https://docs.tryprofound.com/llms.txt
> Use this file to discover all available pages before exploring further.
# Using Looker in Agents
After connecting your Looker account, the following actions become available as Agent steps. Each step requires you to select a connected **Looker Account** from a dropdown.
Profound pulls **query results** from Looker — rows of metrics and dimensions your analytics team has already modeled — not Looker dashboards or visualizations. Results are returned as structured data (typically JSON) that later agent steps can summarize, compare, or combine with other integrations.
## Available actions
| Action | What it does |
| -------------------- | ----------------------------------------------------------------------------------------- |
| **Run Look** | Executes a saved Look by ID and returns the result |
| **Run Inline Query** | Builds and runs a query against a model, Explore, and field list you specify in the agent |
#### **Run Look**
Execute a saved Look from your Looker instance and return the result as structured data.
**Required inputs**
* **Looker Account**
* **Look** — a saved Look from the selected account. The dropdown loads Looks from Looker and shows each Look's title with its ID.
**Optional inputs**
* **Result Format** — the format returned by Looker. Defaults to `json_bi`, which is recommended for downstream agent steps.
* **Row Limit** — maximum number of rows to return.
**Output**
* **Look Results** — a JSON value containing the returned Looker result.
**Tip:** Use **Run Look** when the query already exists as a saved Look in Looker. This keeps business logic governed in Looker and is the safest option for recurring agent automations.
#### **Run Inline Query**
Run a query against a specific Looker model and Explore with fields, filters, sorts, and advanced query options.
**Required inputs**
* **Looker Account**
* **Model** — the Looker model name, such as `marketing`
* **Explore** — the Explore name, such as `web_event`
* **Fields** — comma-separated Looker fields, such as `web_event.date, web_event.unique_users`
**Optional inputs**
* **Sorts** — comma-separated sorts, such as `web_event.date desc`
* **Row Limit** — maximum rows to return. Defaults to `500`.
* **Result Format** — defaults to `json_bi`
**Advanced inputs**
* **Filter Expression** — Looker filter expression using `${field}` syntax
* **Pivots** — comma-separated fields to pivot on
* **Subtotals** — comma-separated fields to subtotal on
* **Fill Fields** — comma-separated fields to fill missing rows for, usually date fields
* **Dynamic Fields** — JSON string for table calculations or custom dimensions
* **Column Limit** — maximum number of columns to return
* **Total** — whether Looker should include totals
* **Row Total** — row total type, such as `right`
* **Query Timezone** — optional IANA timezone, such as `America/Los_Angeles`
* **Server Table Calcs** — whether table calculations should run on the server
* **Apply Formatting** — whether Looker formatting should be applied to the result
**Output**
* **Query Results** — a JSON value containing the returned Looker result.
**Tip:** Use field names exactly as they appear in Looker, including the `view.field` prefix. For date filters, use Looker filter syntax from your LookML model (for example, `30 days` on a date field).
***
## What data is available
Query results depend on the Look or Explore you run. In general, Looker returns the fields defined in your LookML model:
| Data type | Description |
| ---------------------- | -------------------------------------------------------------------------------------- |
| **Dimensions** | Attributes you group or filter by, such as date, channel, region, or product |
| **Measures** | Aggregated metrics, such as revenue, session count, conversion rate, or unique users |
| **Filters & sorts** | Applied at query time — either baked into a saved Look or specified in an inline query |
| **Totals & subtotals** | Optional row or column totals when enabled on inline queries |
Profound returns the raw query result as structured data (typically JSON). Your agent or sheet decides what to do with it — summarize it, compare periods, or pass rows to another step.
## Supported result formats
Both actions support multiple Looker result formats. **`json_bi`** is the default and recommended format for agent workflows because it returns structured JSON that downstream steps can parse reliably.
Other supported formats: `json`, `csv`, `txt`, `html`, `md`, `xlsx`, `sql`, `png`, and `jpg`. Use non-JSON formats only when a later step explicitly needs that output type.
## Limits
* **Run Inline Query** defaults to **500 rows** and supports up to **5,000 rows** per run.
* **Run Look** supports an optional row limit override up to **5,000 rows**.
* The connected Looker user's permissions still apply — Profound cannot return data the API user cannot access.
***
## Choosing an action
| Use **Run Look** when… | Use **Run Inline Query** when… |
| ------------------------------------------------------------- | --------------------------------------------------------- |
| The query is already saved and approved in Looker | The agent needs dynamic fields, filters, or date windows |
| You want Looker to remain the source of truth for query logic | You are building parameterized reporting inside the agent |
| You need a simple, governed recurring report | Earlier agent steps determine which metrics to fetch |
***
## Common use cases
* Pull a saved traffic or revenue Look into a weekly summary agent
* Fetch governed KPIs from Looker and pass them to a content-generation step
* Run dynamic inline queries with date windows that change based on agent inputs
* Combine Looker metrics with data from other integrations in a single agent
* Import a saved Look into a Sheet to analyze metrics alongside agent-generated columns
***
## Best practices
* Prefer `json_bi` for downstream agent steps because it returns structured data
* Start with a saved Look when possible; use inline queries only when the agent needs dynamic logic
* Keep row limits focused so later steps receive only the data they need
* Make sure the connected Looker user has access to every model, Explore, field, and Look used by the agent
***
## Common errors
The Look dropdown depends on the selected Looker account. Choose a **Looker Account** before selecting a saved Look.
The connected Looker user may not have access to any saved Looks, or the Looks may be in folders the user cannot access.
Inline queries must use Looker model, Explore, and field names exactly as they exist in Looker.
Profound uses the connected Looker API user. Update that user's Looker role, model set, or folder permissions if the agent cannot access the requested data.