Fetch MCP Server

MCP Server Configuration

Copy
{
  "mcpServers": {
    "fetch": {
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-fetch"]
    }
  }
}

With Custom User Agent

Copy
{
  "mcpServers": {
    "fetch": {
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-fetch"],
      "env": {
        "USER_AGENT": "Claude-Code-Fetch/1.0"
      }
    }
  }
}

Available Tools

Tool	Description
fetch	Fetch a URL and return content in LLM-friendly format

Tool Parameters

Copy
{
  "url": "string — The URL to fetch",
  "max_length": "integer (optional) — Maximum content length in characters",
  "start_index": "integer (optional) — Start reading from this character position",
  "raw": "boolean (optional) — Return raw content without markdown conversion"
}

Example Usage

# Fetch a documentation page (auto-converts HTML to markdown)
fetch: { url: "https://docs.example.com/api/auth" }

# Fetch API response (JSON)
fetch: { url: "https://api.example.com/status", raw: true }

# Fetch with length limit (for large pages)
fetch: { url: "https://example.com/long-article", max_length: 5000 }

# Paginated reading of long content
fetch: { url: "https://example.com/docs", max_length: 5000, start_index: 0 }
fetch: { url: "https://example.com/docs", max_length: 5000, start_index: 5000 }

Setup

1. Add the configuration to .claude/settings.json
2. No API keys required
3. Requires network access for fetching URLs

Use Cases

• Documentation lookup — Fetch library docs, API references, changelogs
• API responses — Check endpoint responses, verify API behavior
• Content extraction — Pull text from web pages for analysis
• Research — Read online articles, blog posts, release notes

Security Notes

• Respects robots.txt by default
• Follows redirects (up to a reasonable limit)
• Does not execute JavaScript — only fetches static content
• Be mindful of rate limiting on target sites
• Do not use to fetch pages requiring authentication (no cookie/session support)
• Content is processed locally — not stored or forwarded to third parties