In short. Claude API web fetch reads URL content but does not provide documented JavaScript rendering for dynamic websites.
Definition: Claude's API web fetch tool retrieves text from a specified URL, but Anthropic documents that it currently does not support websites dynamically rendered with JavaScript. A page whose relevant content appears only after browser-side execution therefore falls outside the documented web fetch capability.
Key facts
- Anthropic's web fetch tool documentation says the tool retrieves full content from specified web pages and PDF documents.
- The same documentation states that the web fetch tool currently does not support websites dynamically rendered with JavaScript.
- Anthropic recommends its browser use tool for pages that need JavaScript rendering, clicking, or form filling.
- Google explains in its JavaScript SEO basics that an app shell can leave actual content out of the initial HTML until JavaScript runs.
- Google also documents that server-side or pre-rendering helps crawlers because not all bots can run JavaScript.
What does the rendering restriction mean?
JavaScript rendering is the browser step that executes scripts and changes the page after the server returns the initial HTML. A server-side rendered page puts its content in the HTTP response. A client-side app shell can return a document without the actual content and load that content only after JavaScript executes.
Google's documentation makes this distinction explicit: parsing the HTML response works well when the response contains all content, while some JavaScript sites require execution before the generated content becomes visible. This description supplies a useful test for the Anthropic restriction. If the text a reader needs exists only in the post-execution page, the documented web fetch path does not provide the browser rendering required to obtain it.
Anthropic's web fetch reference names browser use as the route for pages that need a real browser. It specifically lists JavaScript rendering, clicking, and filling forms as examples, and describes the browser tool as a client tool in which the application drives the browser and returns page text or screenshots to Claude.
What does web fetch retrieve?
Web fetch is a server tool. Anthropic says that the API fetches during the request and inserts the result into the conversation. When the tool is added to an API request, Claude decides when to fetch, the API retrieves the full text from the specified URL, and Claude analyzes the fetched content.
The current documentation shows this tool declaration:
{
"type": "web_fetch_20260318",
"name": "web_fetch"
}
The result can include optional citations. PDF handling follows a separate documented path: the API returns PDF content as base64-encoded data and processes it like a directly attached PDF document. The rendering restriction concerns dynamically rendered websites, so adding the web fetch tool does not turn it into a general browser session.
What should a website expose for this fetch path?
The practical implication is that important explanatory text should be present in the HTML response when a page is intended to be readable by a text fetcher. Google's documentation says server-side or pre-rendering can make a site faster for users and crawlers, and records that not all bots can run JavaScript. It also says that Google uses rendered HTML for indexing after a headless Chromium step, which describes a capability Anthropic does not document for web fetch.
A mixed page may contain some readable response text and some content added later. Anthropic does not publish a separate rule for how much of such a page web fetch returns, so the documented boundary is the presence or absence of browser-rendered content rather than a stated percentage or token threshold.
Related terms
- JavaScript rendering and AI crawlers
- Server-side rendering
- Claude web search
- How does Claude web fetch differ from web search?
FAQ
Can Claude web fetch click a button that reveals page text?
Anthropic documents clicking as a task for which a real browser is needed and points to browser use. Web fetch itself has the documented JavaScript rendering restriction.
Can the tool read a PDF at a supplied URL?
Yes. Anthropic says the API returns PDF content as base64-encoded data and processes it like a directly attached PDF document.
Does the latest web fetch version add browser rendering?
Anthropic documents newer web fetch versions for capabilities such as dynamic filtering, while the web fetch reference still states that dynamically JavaScript-rendered websites are unsupported. Browser use remains the documented option for pages that need browser execution.
Sources
- Anthropic, "Web fetch tool", undated
- Google Search Central, "Understand the JavaScript SEO basics", 4 March 2026
Check how a page reads to answer engines. Open the GEO score analyzer