Skip to main content
The web_fetch tool allows the Gemini agent to retrieve and process content from specific URLs provided in your prompt.

​
Overview

The agent automatically uses this tool when you include URLs in your prompt and request specific operations like summarization or data extraction.

​
Parameters

​
prompt
string
required
A request containing up to 20 valid URLs (starting with http:// or https://) and instructions on how to process them

​
How It Works

1

URL Detection

The model detects URLs in your prompt and determines if web fetching is needed
2

Confirmation

A confirmation dialog appears showing the URLs that will be fetched
3

Content Retrieval

The tool uses the Gemini API’s urlContext feature for retrieval. If API access fails, it attempts to fetch raw content directly from your local machine.
4

Processing

Returns a synthesized response with source attribution based on your instructions

​
Use Cases

Article Summarization

Summarize technical articles or blog posts

Data Comparison

Compare information between multiple web pages

Information Extraction

Extract specific data from documentation sites

Content Analysis

Analyze and synthesize content from web resources

​
Examples

​
Summarize an Article

Summarize the key points from this article:
https://example.com/technical-article
The model will:
  1. Detect the URL in your prompt
  2. Request confirmation to fetch the URL
  3. Retrieve the article content
  4. Generate a summary of the key points

​
Compare Multiple Sources

Compare the approaches described in these two articles:
https://example.com/article-1
https://example.com/article-2

What are the main differences?
The model will:
  1. Fetch both URLs
  2. Analyze the content from each source
  3. Identify and compare the different approaches
  4. Provide a structured comparison

​
Extract Specific Information

What are the installation requirements mentioned at:
https://example.com/docs/installation
The model will:
  1. Fetch the documentation page
  2. Identify the installation requirements section
  3. Extract and list the specific requirements

​
Limitations

URL Limits: The tool supports up to 20 URLs per request.
Only URLs starting with http:// or https:// are supported. Other protocols (ftp://, file://, etc.) are not processed.

​
Fallback Behavior

If the Gemini API’s urlContext feature is unavailable:
  1. The tool attempts to fetch raw HTML content directly from your machine
  2. Content is processed locally and sent to the model
  3. Some features like JavaScript-rendered content may not be available

​
Response Format

The tool returns synthesized content with:
  • Source Attribution: Each piece of information is attributed to its source URL
  • Structured Content: Content is organized based on your prompt instructions
  • Original Context: The full context is preserved for the model to reason about

​
Security

All web fetches trigger a confirmation dialog showing the exact URLs that will be accessed. Review these carefully before approving.

​
Next Steps

Web Tools Guide

Practical examples of using web fetch

Web Search

Learn about the web search tool for general queries

Build docs developers (and LLMs) love

Get started for freeTalk to us