Skip to main content

What is Browser Automation?

Parallel AI can execute browser tasks described in plain natural language. You submit a task description, receive a task ID, and poll for results when they’re ready. This lets you automate complex web interactions without writing brittle selectors or scripts.

Session Types

Choose the session type that fits your use case:

Submit a Task

Send a POST request to /api/v0/browser-task with your task description and session type.
Response:
Save the taskId from the response — you’ll need it to poll for results or cancel the task.

Poll for Results

Retrieve the current status and result of a task by calling GET /api/v0/browser-task/{task_id}.
Possible status values: Example completed response:
Poll at a reasonable interval (e.g. every 2–5 seconds) to avoid unnecessary API calls. Most tasks complete within 30 seconds.

Cancel a Task

Send a DELETE request to cancel a task. Cancellation is only possible while the task is in pending or running status.
Tasks that have already reached completed, failed, or cancelled status cannot be cancelled again.

Authenticated Sessions

Use authenticated sessions to run tasks inside accounts you’ve connected via Parallel AI integrations (e.g. LinkedIn, Twitter).
1

List available integrations

First, retrieve your saved integrations to find the correct integrationId:
2

Submit with an integrationId

Pass sessionType: "authenticated" and the integrationId in your task request:

Residential Proxy

Route your browser session through a residential IP address. Provide a US ZIP code to target a specific location:
Residential sessions are useful for tasks that may be geo-restricted or that require a non-datacenter IP to avoid bot detection.

Vision Mode

Enable useVision: true to switch to screenshot-driven automation. In vision mode, the browser agent perceives pages as images rather than DOM elements — useful for visually complex or canvas-based interfaces.
For the full Browser Automation API reference, see Submit a Browser Task.