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

# Browser Automation

> Automate interactions with web applications from a workflow node, driving a real browser with natural-language task instructions.

<Badge color="blue" icon="cloud">SaaS · 5.13</Badge>

<Info>
  **Available starting with FlowX.AI 5.13.0** on SaaS deployments. Self-hosted deployments will receive it with the next LTS release family.
</Info>

## Overview

The **Browser Automation** node drives a real browser to complete tasks on web applications that have no API: filling multi-page forms, navigating portals, and retrieving results. You describe the task in natural language, and an AI agent plans and executes the browser interactions, wrapped in the platform's execution, logging, and document storage layers.

<CardGroup cols={2}>
  <Card title="Natural-language tasks" icon="message">
    Describe what to do on the target site; the agent plans the browser steps
  </Card>

  <Card title="Verified outcomes" icon="badge-check">
    Results are checked against the target site's recorded state changes, not the agent's own success claim
  </Card>

  <Card title="Persistent sessions" icon="window-restore">
    Multi-step automations continue in the same browser session across workflow steps
  </Card>

  <Card title="Full audit trail" icon="list-check">
    Every run is logged through the platform's execution and observability layers
  </Card>
</CardGroup>

***

## When to use it

Use Browser Automation when the system you need to reach exposes no API or integration point: legacy portals, third-party sites, and internal tools where the browser is the only interface. When the target exposes an API, prefer a [RESTful System integration](../../docs/platform-deep-dive/integrations/integration-designer); when you only need page content, use the [Web Page Extractor](./web-page-extractor) node.

***

## Reliability

The node's execution approach is grounded in FlowX.AI's published browser-agent benchmark: 405 verified runs across nine models and nine controlled web tasks, showing that outcome-state verification catches failures that agent self-reports miss. See the [research page](../research#browser-agent-reliability-benchmark) and the [full benchmark](https://www.flowx.ai/blog/browser-agents-dont-fail-on-capability-they-fail-on-consistency).

***

## Related resources

<CardGroup cols={2}>
  <Card title="Web Page Extractor" icon="globe" href="./web-page-extractor">
    Read-only content collection from web pages
  </Card>

  <Card title="Browser-agent reliability benchmark" icon="flask" href="../research#browser-agent-reliability-benchmark">
    The 405-run benchmark behind the node's verification approach
  </Card>
</CardGroup>


## Related topics

- [Web Crawler setup](/5.9/setup-guides/web-crawler-setup.md)
- [Custom Agent node](/5.9/docs/platform-deep-dive/integrations/custom-agent-node.md)
- [Agent Builder use cases](/5.9/ai-platform/agent-builder/use-cases.md)
- [Microservices setup guides](/5.9/setup-guides/setup-guides-overview.md)
- [Data mappers](/5.9/docs/building-blocks/process/data-mappers.md)