Browserbase icon

Browserbase

Browser automation, web search, and page fetches with Browserbase.

Actions7

Fetch → Run an Agent

AI-generated

Overview

This node runs an AI agent to perform browser automation tasks starting from a specified URL. It uses a selected AI model to navigate and interact with web pages based on given instructions, supporting different interaction modes such as vision-based (CUA), DOM selector-based, or a hybrid approach. The node is useful for automating complex web tasks like data extraction, navigation, and interaction with dynamic web content.

Use Case Examples

  1. Automate data extraction from a pricing page by instructing the agent to find and list all plan names and prices.
  2. Navigate through a multi-step web form and submit information automatically using the agent's reasoning and navigation capabilities.
  3. Perform automated testing of web interfaces by instructing the agent to interact with UI elements and verify expected outcomes.

Properties

Name Meaning
Model The AI model that drives the browser and runs the agent, handling navigation and reasoning. Different models can be selected based on provider and capabilities.
Starting URL The initial web page URL where the agent begins its task.
Instruction The specific task or instructions for the agent to complete on the web page.
Model Options Optional settings to override the reasoning model based on the interaction mode (CUA, DOM, Hybrid), cursor highlighting, max steps, and custom system prompt.
Variables Sensitive data passed to the agent as variables with placeholders and descriptions, never exposing actual values to the AI.
Browser Options Settings for browser behavior during the session, such as blocking ads, logging, recording, solving captchas, viewport size, and verified identity.
Session Options Options for session management including context reuse, keep alive, persistence, region, timeout, proxy usage, and user metadata.

Output

JSON

  • success - Indicates if the agent task was successful.
  • message - Message describing the result or status of the task.
  • actions - List of actions performed by the agent during execution.
  • completed - Boolean indicating if the task was completed.
  • usage - Usage statistics related to the agent execution.
  • sessionId - Identifier of the browser session used.
  • contextId - Optional context ID if session context was reused.

Dependencies

  • Browserbase API key credential

Troubleshooting

  • Ensure the selected Agent Model supports the chosen interaction mode (CUA, DOM, Hybrid). For example, CUA mode requires a computer-use-capable model.
  • If using a user-provided API key, ensure the Model and Agent Model are from the same provider to avoid authentication errors.
  • Check that the Starting URL is a valid and accessible web address.
  • Verify that session options like context ID and proxies are correctly configured to avoid session failures.
  • Common error messages include API call failures with details; ensure network connectivity and valid API keys.

Links

Discussion