Browser

Browser actions drive a headless Chromium instance: open a page, click an element, then read the rendered HTML or take a screenshot. Use them for pages that only make sense once JavaScript has run, where an HTTP request would return an empty shell.

All four actions need a chromium integration. See Configuration Reference for setup details.

The integration holds one browser session, shared by every action that uses it — across steps and across concurrent requests. chromium/body, chromium/click, and chromium/screenshot all act on whatever page chromium/navigate last opened, so keep these steps in order, don’t put them in a parallel group, and expect interference if two requests drive the same integration at once. Give a workflow that needs isolation its own Chromium integration.

chromium/navigate

Navigates the browser to a URL.

The Chromium session to drive.


YAML Key	`integration`
Type	integration
Required	Yes

URL

The address to open.


YAML Key	`url`
Type	string
Required	Yes

Output

{{ .step_id }} is an object:

Path	Type	Description
`success`	boolean	Whether the page was reached
`url`	string	The address that was opened
`message`	string	A sentence describing what happened

Example

actions:
  open_page:
    type: chromium/navigate
    config:
      integration: my_chromium
      url: "https://example.com/reports/{{ param \"id\" }}"
    next: action.read_page
    fail: response.error

chromium/click

Clicks an element on the current page using a CSS selector.

Browser

The Chromium session to drive.


YAML Key	`integration`
Type	integration
Required	Yes

Selector

The CSS selector of the element to click.


YAML Key	`selector`
Type	string
Required	Yes

For example #submit-button, .my-class, or button[type='submit'].

Output

{{ .step_id }} is an object:

Path	Type	Description
`success`	boolean	Whether the element was clicked
`selector`	string	The selector that was clicked

Example

actions:
  expand_details:
    type: chromium/click
    config:
      integration: my_chromium
      selector: "button.show-more"
    next: action.read_page

chromium/body

Gets the rendered body HTML of the current page.

Browser

The Chromium session to drive.


YAML Key	`integration`
Type	integration
Required	Yes

Output

{{ .step_id }} holds the page’s rendered HTML.

Example

actions:
  read_page:
    type: chromium/body
    config:
      integration: my_chromium
    next: action.extract

chromium/screenshot

Takes a full-page screenshot and returns it as a PNG.

Browser

The Chromium session to drive.


YAML Key	`integration`
Type	integration
Required	Yes

Output

{{ .step_id }} holds the screenshot as PNG bytes. It is a file, not text, so pass it to a step that takes a file — download to save it, or an agent’s fileUpload to show it to a model.

Example

Capture a page and save it to the workspace:

actions:
  capture:
    type: chromium/screenshot
    config:
      integration: my_chromium
    next: action.save_shot

  save_shot:
    type: download
    config:
      file:
        type: action
        identifier: capture
      destinationPath: "screenshots"
      fileName: "{{ param \"id\" }}.png"
      overwrite: true
    next: response.success

Common Patterns

Scrape a JavaScript-Rendered Page

Open the page, expand it, then read the rendered HTML:

actions:
  open_page:
    type: chromium/navigate
    config:
      integration: my_chromium
      url: "{{ param \"url\" }}"
    next: action.expand

  expand:
    type: chromium/click
    config:
      integration: my_chromium
      selector: "button.show-more"
    next: action.read_body

  read_body:
    type: chromium/body
    config:
      integration: my_chromium
    next: action.summarize

  summarize:
    type: agent
    config:
      providerID: 2
      systemPrompt: "Extract the report's key figures as JSON."
      userPrompt: "{{ .read_body }}"
    next: response.summary

Next Steps

HTTP Requests

Fetch a page without a browser when JavaScript isn’t needed.

System

Save captured files to the workspace.

Transformation

Parse scraped HTML with JavaScript.

Actions Overview

Learn the fundamentals of ServFlow actions.

Getting Started

Quickstart

Running ServFlow

Concepts

References

Browser

chromium/navigate

Browser

URL

Output

Example

chromium/click

Browser

Selector

Output

Example

chromium/body

Browser

Output

Example

chromium/screenshot

Browser

Output

Example

Common Patterns

Scrape a JavaScript-Rendered Page

Next Steps

HTTP Requests

System

Transformation

Actions Overview

​chromium/navigate

​Browser

​URL

​Output

​Example

​chromium/click

​Browser

​Selector

​Output

​Example

​chromium/body

​Browser

​Output

​Example

​chromium/screenshot

​Browser

​Output

​Example

​Common Patterns

​Scrape a JavaScript-Rendered Page

​Next Steps

HTTP Requests

System

Transformation

Actions Overview

chromium/navigate

Browser

URL

Output

Example

chromium/click

Browser

Selector

Output

Example

chromium/body

Browser

Output

Example

chromium/screenshot

Browser

Output

Example

Common Patterns

Scrape a JavaScript-Rendered Page

Next Steps