How can Claude web fetch retrieve a fresh page instead of cached content?

The Citead team, updated

In short. Set use_cache to false with web_fetch_20260309 or later when a request needs fresh content.

Definition: Claude web fetch is Anthropic's server tool for retrieving the content of a specified web page or PDF. As of 17 October 2026, Anthropic's web fetch tool documentation says that use_cache: false bypasses the tool cache on web_fetch_20260309 and later versions.

Key facts

What setting requests fresh content?

The setting belongs in the web_fetch tool configuration in the Messages API request. Anthropic documents true as the default and false as the value that bypasses the web fetch cache. The request can use the following shape:

{
  "tools": [
    {
      "type": "web_fetch_20260309",
      "name": "web_fetch",
      "use_cache": false
    }
  ]
}

The same documentation says that the cache is managed automatically and may change over time for different content types and usage patterns. It also says that returned content may not always reflect the latest version available at the URL. The false value addresses the web fetch tool's cache choice for that request; the documentation does not give a cache lifetime.

Which web fetch versions support use_cache?

Anthropic lists web_fetch_20260318 as the latest web fetch version in the documentation used for this entry. It describes web_fetch_20260309 as the previous version with dynamic filtering and cache bypass, web_fetch_20260209 as a version with dynamic filtering only, and web_fetch_20250910 as the basic fetch version. The page states that use_cache requires web_fetch_20260309 or later, including web_fetch_20260318.

This version choice also depends on the deployment described by Anthropic. Web fetch is available on the Claude API, Claude Platform on AWS, and Microsoft Foundry. Microsoft Foundry deployments hosted on Azure support only the basic web_fetch_20250910 tool, while deployments hosted on Anthropic support all versions. The documentation lists web fetch as unavailable on Amazon Bedrock and Google Cloud.

What can still affect the fetched page?

The cache setting controls whether Anthropic's web fetch cache may supply the result. A fresh fetch still follows the tool's other documented limits. Web fetch currently does not support websites dynamically rendered with JavaScript. The max_content_tokens parameter can truncate text when the fetched content exceeds its limit, and the limit applies to text rather than binary content such as PDFs.

Fetch results include the fetched url, a document content block, and retrieved_at, a timestamp for when the content was retrieved. The API executes web fetch as a server tool, so the application sees the tool call and result without handling execution or returning a client-side tool_result.

FAQ

Is use_cache false by default?

No. Anthropic documents true as the default. Set use_cache to false when the request explicitly needs fresh content or the source changes rapidly.

Can web_fetch_20250910 use use_cache?

Anthropic's documentation requires web_fetch_20260309 or later for use_cache. The page identifies web_fetch_20250910 as the basic fetch version and shows the cache parameter only with the later version.

Does cache bypass make a page usable when its content needs JavaScript?

No. The web fetch tool documentation says that it does not support websites dynamically rendered with JavaScript. A cache bypass changes cache handling, while JavaScript rendering is a separate documented limitation.

Sources

  1. Anthropic, "Web fetch tool", accessed 17 October 2026
  2. Anthropic, "Server tools", accessed 17 October 2026

Check how a page reads to answer engines. Open the GEO score analyzer

Test a page and see what to fix first

The analysis usually takes less than 10 seconds and needs no account.

400,000+ pages scored in the public benchmark