Skip to main content
Claude Code includes two web tools: WebFetch for fetching a specific URL and WebSearch for discovering information across the web.

​
WebFetch

Fetches content from a URL and processes it with an AI model to extract the information you need. HTML is converted to Markdown before processing.

​
Parameters

​
url
string
required
The fully-formed URL to fetch. HTTP URLs are automatically upgraded to HTTPS. Must be a valid URL — an invalid URL will produce an error without making a network request.
​
prompt
string
required
Instructions describing what information to extract from the fetched page. The content is processed by a small, fast model using this prompt, and the model’s response is returned.Example: "What are the required parameters for the createUser endpoint?"

​
Output

​
result
string
The processed response from the model, based on the fetched content and the prompt.
​
url
string
The URL that was fetched.
​
code
number
HTTP response status code.
​
codeText
string
HTTP response status text (e.g., "OK", "Not Found").
​
bytes
number
Size of the fetched content in bytes.
​
durationMs
number
Time taken to fetch and process the content in milliseconds.

​
Usage notes

  • Authenticated URLs will fail. WebFetch cannot access Google Docs, Confluence, Jira, GitHub private repos, or any URL requiring a login. Use a specialized MCP tool if you need authenticated access.
  • If an MCP server provides a web fetch tool, prefer it — it may have fewer restrictions.
  • For GitHub URLs, prefer using the gh CLI via the Bash tool (e.g., gh pr view, gh issue view).
  • HTTP redirects to the same host are followed automatically. Redirects to a different host are surfaced in the result field with the redirect URL. Make a new WebFetch call with the redirect URL to continue.
  • Results may be summarized when content is very large.
  • A 15-minute cache is maintained for repeated fetches of the same URL.
  • Binary content (PDFs, etc.) is saved to a temporary file and the path is noted in result.

​
Permissions

WebFetch requires permission before fetching from a domain. Claude Code will prompt you on the first request to a new domain. You can permanently allow a domain in your settings.

​
Examples

Fetch API documentation
url: "https://docs.example.com/api/reference"
prompt: "List all available endpoints and their HTTP methods"
Extract a code example
url: "https://docs.example.com/quickstart"
prompt: "Show me the minimal code example for getting started"
Check library version requirements
url: "https://github.com/example/library/blob/main/package.json"
prompt: "What are the peer dependencies and their version constraints?"

​
WebSearch

Searches the web and returns results with summaries and source links. Uses Anthropic’s web search capability, which performs searches automatically within a single API call.
WebSearch is only available when using Anthropic’s API directly (first-party), Vertex AI with Claude 4.0+ models, or Foundry. It is not available with all API configurations.

​
Parameters

​
query
string
required
The search query. Minimum 2 characters. Be specific — include the current year when searching for recent information, documentation, or news.
​
allowed_domains
string[]
When provided, only include search results from these domains. Cannot be used together with blocked_domains.Example: ["docs.python.org", "pypi.org"]
​
blocked_domains
string[]
When provided, exclude search results from these domains. Cannot be used together with allowed_domains.Example: ["stackoverflow.com"]

​
Output

​
query
string
The search query that was executed.
​
results
array
An array of search result objects and text commentary from the model. Each element is either:Or a plain string containing the model’s commentary or summary.
​
durationSeconds
number
Time taken to complete the search operation.

​
Response format

Claude Code’s response to a WebSearch call includes both the search results and a synthesized answer. Sources must be cited at the end of the response as markdown hyperlinks.
Example response format
[Answer based on search results]

Sources:
- [Source Title 1](https://example.com/1)
- [Source Title 2](https://example.com/2)

​
Usage notes

  • WebSearch is only available in the US.
  • A single WebSearch call may internally perform up to 8 sub-searches.
  • Domain filtering (allowed_domains and blocked_domains) is mutually exclusive — only provide one or the other.
  • Always include the current year in queries about recent releases, documentation versions, or current events.

​
Examples

Search for library documentation
query: "React 19 new features 2025"
Search within specific domains
query: "Python asyncio tutorial"
allowed_domains: ["docs.python.org", "realpython.com"]
Search excluding specific domains
query: "best practices for JWT authentication"
blocked_domains: ["medium.com"]