Web tools give Claude the ability to retrieve information from the internet — fetching documentation, searching for solutions, and reading external resources.
Fetches the content of a URL and returns it as text or markdown. Useful for reading documentation, API specs, or any web page.
Usage pattern:
> Fetch the React hooks documentation and explain useCallback
> Get the latest changelog from the express npm page
> Read this Stack Overflow answer: https://...
- Web pages (returned as markdown)
- Plain text files
- HTML pages (converted to readable text)
- JSON APIs
WebFetchTool requires network permission. On first use, Claude will ask for your approval to make web requests.
> Search for how to configure Jest with TypeScript
> Find the documentation for the Zod v4 API
> Look up the error "ECONNREFUSED 127.0.0.1:5432"
- Looking up unfamiliar libraries or APIs
- Finding solutions to error messages
- Checking current best practices
- Locating documentation for dependencies
Combining with WebFetchTool:
Claude often uses WebSearchTool to find relevant URLs, then WebFetchTool to read the full content:
- Search for “Prisma soft delete pattern”
- Identify the most relevant result
- Fetch the full page for detailed guidance
You can tell Claude to search the web explicitly: “Search the web for the best way to handle JWT refresh tokens in Express.” Claude will use WebSearchTool automatically when it needs up-to-date information.