Web Access icon

Web Access

Use the Web Access tool to perform any task on given URL(s). Just describe what you want in natural language and the tool will: 1. Scrape the webpage content automatically 2. Try pattern matching first (fast, free) 3. Use LLM agent if needed (for complex tasks) EXAMPLES: - "Find the contact email" → Extracts email addresses - "Research this company" → Gathers and synthesizes company info - "Get all product prices" → Extracts product listings - "What services do they offer?" → Analyzes and summarizes content - "Extract the main article text" → Gets readable text content RETURNS: { text: "result", sources: ["urls"] } with the answer to your task.

Web Access

AI-generated

Summary

Perform a specified natural language task on one or multiple webpages by providing their URLs, optionally using pattern matching or a large language model (LLM) for complex data extraction and synthesis.

Inputs

  • URLs (required) — One or more full URLs (including https://) on which to perform the task. The node operates on each URL provided.
  • Task (required) — A natural language description of the desired task to perform on the URLs, such as extracting contact emails, researching a company, listing products with prices, or summarizing content.
  • LLM Provider (required) — Choice of using no language model (pattern matching only for fast, cost-free extraction) or an OpenAI-compatible LLM for more complex understanding and synthesis.
  • Model — Specifies the model to use when LLM is enabled, selectable from a list or by ID, such as 'openai/gpt-4o-mini'.

Output shape

A single item containing a JSON object for each URL processed, including properties: text (the extracted or synthesized result), sources (the processed URLs), success (boolean), meta (metadata about scraping and LLM usage), and error (if any).

Returns combined results for all input URLs. Handles multiple URLs by performing the task on each individually. If pattern matching extraction fails and LLM is enabled, uses the agent to attempt task completion. Errors per URL are included in the results. Supports pagination only by providing additional URLs as inputs.

Examples

Example 1: Extract contact emails from multiple company websites.

Specify the URLs list with company homepage URLs, set Task to 'Find contact email', choose LLM Provider as 'None (Pattern Matching Only)' for quick extraction.

Example 2: Research detailed company information from a webpage.

Provide the company page URL in URLs, set Task to 'Research this company in detail', enable LLM Provider with an OpenAI-compatible model selected for deep analysis.

Discussion