Browser Use Block

The Browser Use block executes browser automation tasks with BrowserUse, letting an AI agent navigate the web, scrape data, and perform actions as if a real user were driving the browser. Reach for it when a workflow needs live, interactive web behavior that a simple HTTP request cannot accomplish. The task runs asynchronously and the block polls for completion before returning results.

Overview

When to Use

• Automate multi-step web interactions (logging in, filling forms, clicking through flows) that require a real browser.
• Scrape data from sites that render content dynamically with JavaScript.
• Perform research or data collection tasks described in natural language rather than hardcoded selectors.
• Drive site actions that depend on session state or browser cookies (enable Save Browser Data to persist the session).
• Run tasks that must reason about page content step by step using a chosen AI model.
• Integrate browser-based automation into a larger workflow that then processes the extracted output or steps downstream.

Configuration

Task

The natural-language instruction describing what the browser agent should do. This is the primary prompt sent to the BrowserUse AI agent. Required.

Example: Go to https://example.com, search for "AI workflows", and return the top 5 result titles.

Variables (Secrets)

A key/value table of variables passed into the task as secrets (e.g. credentials, tokens, or dynamic values). Each row has a Key and a Value column. The tool sends these as a secrets object so the agent can use them without embedding raw credentials in the task text.

Model

Dropdown selecting the AI model that drives the browser agent. Default is gpt-4o.

Save Browser Data

Switch (boolean) to persist browser session data (cookies, local storage) from the run so subsequent tasks can reuse the authenticated session.

API Key

Your BrowserUse cloud API key, stored as a password field. Obtain one from the BrowserUse dashboard. Required. Use {{BROWSER_USE_API_KEY}} to pull the value from a workspace secret rather than hard-coding it.

Inputs & Outputs

• Inputs:

  • task (string) — Browser automation task description
  • apiKey (string) — BrowserUse API key
  • variables (json) — Task variables / secrets passed to the agent
  • model (string) — AI model to use for the browser agent
  • save_browser_data (boolean) — Whether to save browser session data

• Outputs:

  • id (string) — Task execution identifier assigned by BrowserUse
  • success (boolean) — Whether the task completed successfully (finished status)
  • output (json) — Final output data returned by the task
  • steps (json) — Array of execution steps taken by the agent, each containing goal, evaluation, URL, and any extracted data

Tools

Browser Use (browser_use_run_task) — Submits a browser automation task to the BrowserUse API (https://api.browser-use.com/api/v1/run-task), then polls the task status endpoint every 5 seconds (up to 3 minutes) until the task reaches finished, failed, or stopped. Returns the final task output and step-by-step execution trace. Ad blocking and element highlighting are enabled by default on every run.

YAML Example

browser_use_1:
  type: browser_use
  name: "Browser Use"
  inputs:
    task: "Go to https://example.com and extract the page title and first paragraph"
    model: "gpt-4o"
    save_browser_data: false
    apiKey: "{{BROWSER_USE_API_KEY}}"
    variables:
      - Key: "username"
        Value: "{{SITE_USERNAME}}"
      - Key: "password"
        Value: "{{SITE_PASSWORD}}"
  connections:
    outgoing:
      - target: next-block-id