> ## Documentation Index
> Fetch the complete documentation index at: https://docs.ux.report/llms.txt
> Use this file to discover all available pages before exploring further.

# Agents

> Templates and a linear canvas that turn Opportunities into drafts and applied work.

**Agents** run the work — they don’t stop at a brief. Start from a template or
a blank canvas, ground on your Knowledge base, then land on an Opportunity
(and optionally Slack).

In the sidebar: **Optimize → Agents** (alpha).

## What Agents are for

* Draft page fixes and FAQ blocks so answer engines can cite you
* Write a blog article from a monitoring signal + a competitor URL
* Pull live page, citation, and visibility data into a run
* Hand finished work to an Opportunity for placement and Apply

Agents are **alpha**. Templates and the canvas are live; CMS publish and
one-click “build an agent from this Opportunity” are still landing.

## Live templates

| Template                             | What it does                                      |
| ------------------------------------ | ------------------------------------------------- |
| **Content optimization suggestions** | One page + one buyer question → concrete fixes    |
| **Optimize this URL**                | Thinner scrape → draft → Opportunity on one URL   |
| **Write a blog article**             | Article draft from topic + competitor URL         |
| **FAQ for AI answers**               | FAQ blocks tuned so engines can cite your product |
| **Starter — blank Opportunity**      | Empty Start → End set to Opportunity              |

Coming next: competitor-gap brief, content refresh, and blog-pillar inject.
Those stay labeled until they run for real.

## Typical flow

1. From an **Opportunity**, press **Run with Agent** — that builds a ready-to-run Agent from the plan (page + buyer question). Or open **Optimize → Agents** and pick a template.
2. Confirm the Start inputs.
3. **Run**. Steps light up as scrape, scorecard, LLM, and delivery finish.
4. Review the output. Execute handoff writes back to the **same Opportunity**. Hub templates create a new one. Slack delivery is optional on the End node.

Fill [Knowledge base](/product/knowledge) before you generate. Brand Model is
required for content drafts. Company files are retrieved by default; an empty
corpus warns and still runs.

## Canvas

The canvas is linear: **Start → middle nodes → End**. You can insert or remove
middle steps (scrape, scorecard, **Visibility Score**, web search, prompt
probe, LLM, conditions).

**Visibility Score** pulls the same Monitoring appearance / share-of-voice
numbers you see on Overview. Pick a metric, a date range (with timezone), and
optional engine / region / competitor filters. It does not invent a composite
score. Downstream nodes can read `{{visibility_snapshot}}` (JSON) or
`{{visibility_score}}` (the number).

Ends today:

| End                     | Meaning                                        |
| ----------------------- | ---------------------------------------------- |
| **Opportunity**         | Writes a work item you can place and Apply     |
| **Slack + Opportunity** | Same, plus a Slack message when connected      |
| **Text**                | Draft you can copy (blog article)              |
| **Deploy / Notify**     | Labeled soft — not a finished publish path yet |

Apply for a page fix still runs from the **Opportunity** after you approve
placement — not from a Deploy button on the canvas.

## Related

* [Knowledge base](/product/knowledge)
* [Opportunities](/product/improve)
* [AI Visibility](/product/ai-visibility)
* [Product overview](/product/overview)