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

# Give your agent actions

> Choose how your agent collects information, calls your API, or works with Google Sheets, then test and review the results.

Actions let your agent work with information during a conversation: connect to external systems, extract structured data, or read and write approved spreadsheet data. Open **Configuration → Actions** to create and manage them.

Choose the action type that fits where your information lives and what your agent needs to do. The linked guides include practical examples you can adapt to your own business.

## Choose an action type

<Columns cols={2}>
  <Card title="Connect a system" icon="plug" href="/product-guides/ai-agent/custom-tools">
    Connect your agent to an external system through an API. Use information from the conversation to fetch up-to-date data or perform operations your API supports.
  </Card>

  <Card title="Collect data" icon="list-check" href="/product-guides/ai-agent/collect-data">
    Extract structured data from a conversation using fields you define. Capture information already shared, ask for missing details, and save the results in Visito.
  </Card>

  <Card title="Google Sheets" icon="table" href="/product-guides/integrations/google-sheets">
    Let your agent read, add, or update spreadsheet rows during a conversation. Choose which columns it can access and what it can do with each one.
  </Card>
</Columns>

| Your goal | Start with |
| - | - |
| Fetch current information or perform an operation through an API | Connect a system |
| Turn conversation details into structured fields saved in Visito | Collect data |
| Read, add, or update data in a spreadsheet | Google Sheets |

A connection and an action serve different purposes. Manage provider accounts under **Connections → Integrations**. Configure what the agent may do under **Actions**. Supported connected integrations can also expose their available capabilities in the Actions list.

## Google Sheets examples

The **Google Sheets** subsection contains the setup guide and its worked examples together.

<Card title="Hairdresser appointments" icon="calendar" href="/product-guides/integrations/google-sheets-hairdresser">
  Prepare availability and bookings tabs, choose columns and permissions, book through Webchat, and check the saved row and recalculated availability.
</Card>

## Set up and test an action

1. **Define its purpose.** Explain when to use it, what information to ask for, and when to confirm with the customer.
2. **Choose its business scope.** Make the action available only where its data and behavior belong.
3. **Configure its inputs or permissions.** Collection actions use fields; API actions use an input schema and endpoint; Sheets actions use column permissions and rules for identifying unique rows.
4. **Choose its mode and test destination.** Follow the action-specific guide before enabling live channel use. New collection actions and Sheets writes start in Playground only. Sheets writes require a separate matching test spreadsheet; read-only Sheets actions use the selected spreadsheet in Playground.
5. **Verify the result.** Check the saved data or API response, not only the agent's final message.

For a controlled Webchat test, the action must be available to that channel and business. Playground and Webchat can differ in where data is saved and whether external delivery occurs; each guide explains the relevant behavior.

## Review what happened

Open **Actions → History** to filter executions by action, environment, and status. Open an execution to inspect its inputs, result, timing, and available technical details. Use **Open conversation** to see the customer's request and the action receipt in context.

For **Collect data**, the conversation's **CRM → Action submissions** panel shows live saved submissions. For **Connect a system**, compare the stored request information with the response; saved request data may not contain the entire transmitted HTTP body. For **Google Sheets**, check the execution result and the relevant spreadsheet cells.

A completed request does not always mean the intended operation succeeded. Check the result returned by the system or the saved data before confirming the outcome to the customer.

<Tip>
  Begin with one small use case and synthetic data. Once the setup and conversation work, review scope and permissions before making it available to customers.
</Tip>

Developers can also manage supported actions and inspect executions through the [Actions API](/api-docs/actions-api).


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.