Hyperbrowser icon

Hyperbrowser

Interact with websites using Hyperbrowser
Join our community

Actions6

Overview

The node "Hyperbrowser" enables interaction with websites through various automated browsing and data extraction operations powered by the Hyperbrowser SDK. It supports tasks such as scraping web pages, crawling multiple pages, extracting structured data, and controlling browser actions via AI agents including Claude and OpenAI.

This node is beneficial in scenarios where you want to automate complex web interactions or gather data from websites without manual browsing. For example:

  • Automatically filling out forms or clicking buttons on a website using AI instructions.
  • Scraping product information or prices from e-commerce sites.
  • Crawling a website to collect content from multiple linked pages.
  • Using AI agents to control computer/browser actions based on natural language tasks.

The "Claude Computer Use" operation specifically uses the Claude AI agent to perform computer control tasks described by user instructions, automating sequences of actions on the browser or system.

Properties

Name Meaning
Task Instructions for browser automation, e.g., "Click the login button and fill in the form".
Options Collection of optional parameters:
- Max Steps Maximum number of steps the agent should take to complete the task (default 25).
- Maximum Pages Maximum number of pages to crawl (used in crawl operation, default 10).
- Only Main Content Whether to return only the main content of the page (boolean, default true).
- Output Format Format of output content: HTML, Links, or Markdown (default Markdown).
- Proxy Country Country code for proxy server if proxy is used (e.g., "US").
- Solve CAPTCHAs Whether to solve CAPTCHAs during scraping (boolean, default false).
- Timeout (Ms) Maximum timeout in milliseconds for navigating to a page (default 15000 ms).
- Use Proxy Whether to use a proxy server for scraping (boolean, default false).
- Use Vision Whether to enable vision capabilities for Browser Use LLM (boolean, default true, browserUse only).

Output

The node outputs an array of JSON objects corresponding to each input item processed. The structure depends on the selected operation:

  • Claude Computer Use:

    • actions: The final result of the AI agent's performed actions as a sequence or description.
    • status: Status code or message indicating success or failure.

  • Other operations (for context):

    • scrape: Contains url, content (page content in requested format), and status.
    • crawl: Contains url, data (crawled content), and status.
    • extract: Contains url, extractedData (structured data per schema), and status.
    • browserUse and openaiCua: Similar to Claude Computer Use but with respective AI agents.

No binary data output is indicated; all outputs are JSON-formatted data describing results or errors.

Dependencies

  • Requires an API key credential for the Hyperbrowser service to authenticate requests.
  • Uses the @hyperbrowser/sdk package to interact with the Hyperbrowser API.
  • Node expects network access to the Hyperbrowser API endpoints.
  • Optional proxy configuration can be set via node properties.

Troubleshooting

  • Common Issues:

    • Invalid or missing API key credential will cause authentication failures.
    • Network timeouts if the target URL is slow or unreachable; adjust the timeout property accordingly.
    • If the AI agent cannot complete the task within the max steps limit, consider increasing it.
    • CAPTCHA challenges may block scraping unless "Solve CAPTCHAs" is enabled.
    • Proxy misconfiguration can lead to connection errors.

  • Error Messages:

    • "Operation \"<operation>\" is not supported": Occurs if an unsupported operation value is provided; verify operation selection.
    • Errors returned from the Hyperbrowser API will be included in the output JSON under an error field if "Continue On Fail" is enabled.

To resolve errors, ensure correct credentials, valid URLs, appropriate options, and stable network connectivity.

Links and References

Discussion