# Record a how-to guide while you click | User Guide | Shotomatic

> Start an Action Capture from Claude or Codex, click through the task yourself, and let the agent write a title and description for every step.

- Type: Public web page
- Canonical page: [https://www.shotomatic.com/guide/ai-agents/recipes/action-capture-guide](https://www.shotomatic.com/guide/ai-agents/recipes/action-capture-guide)
- Markdown page: [https://www.shotomatic.com/guide/ai-agents/recipes/action-capture-guide.md](https://www.shotomatic.com/guide/ai-agents/recipes/action-capture-guide.md)

Browse the user guide

On this page

1.  [Before you start](#before-you-start)
2.  [Ask your agent](#ask-your-agent)
3.  [What happens](#what-happens)
4.  [What Shotomatic asks you](#what-shotomatic-asks-you)
5.  [Finish the guide](#finish-the-guide)
6.  [Limits](#limits)

An agent can start an [Action Capture](https://www.shotomatic.com/guide/screenshot-capture/action-capture) for you. Each click you make in the chosen window saves a step with a click marker. When you finish, Shotomatic builds a Document, and the agent writes a title and a short description for every step.

Shotomatic does not click for the agent. You perform the clicks, or an agent with its own computer-use ability does.

## Before you start

-   [Connect an agent](https://www.shotomatic.com/guide/ai-agents/connect-an-agent).
-   On a Mac, grant **Screen Recording** and **Accessibility**.
-   Open the app or web page you want to document and get it to the starting point.
-   Keep Shotomatic's main window open; an Action Capture cannot start without it.

## Ask your agent

> "Record a guide for exporting a report from this app. I will click through it; stop when I say done, then write a title and a one-line description for each step."

Wait until the agent says the capture has started, then click through the task at a normal pace.

## What happens

1.  The agent finds the window and starts an Action Capture.
2.  Every click in that window saves one screenshot as a step. If a screen changes without a click, ask the agent to save a step now.
3.  When you tell the agent you are done, it stops the capture. Shotomatic builds a Document from the steps.
4.  The agent looks at the steps and writes each Page's title and description. It can also trim the window edges or reorder Pages.

## What Shotomatic asks you

Nothing. Recording, writing step text, and exporting need no confirmation. Shotomatic asks only if the agent later wants to delete something.

## Finish the guide

> "Blur the email address on step 3 and export the guide as a PDF."

Blur and some other annotations need Pro, for an agent as for you. You can also open the Document in Shotomatic and finish it yourself; if the agent edits a Document while it is open in the editor, the editor reloads it.

## Limits

-   On the free plan, an Action Capture ends by itself after 5 steps. See [Limits and safety](https://www.shotomatic.com/guide/ai-agents/limits-and-safety#plan-limits-apply).
-   Only one capture runs at a time, whether you or an agent started it.

Related: [Documents editor](https://www.shotomatic.com/guide/documents/editor), [Annotations and framing](https://www.shotomatic.com/guide/documents/annotations-and-framing).

## On this page

1.  [Before you start](#before-you-start)
2.  [Ask your agent](#ask-your-agent)
3.  [What happens](#what-happens)
4.  [What Shotomatic asks you](#what-shotomatic-asks-you)
5.  [Finish the guide](#finish-the-guide)
6.  [Limits](#limits)

## About this version

This Markdown page is generated from the server-rendered public page on Shotomatic. Interactive controls and site navigation are omitted. Use the canonical page when citing or sharing it with people.
