Gives Claude three tools for working with web content through Google's Gemini API. You can summarize pages, compare content across multiple URLs, or extract specific information using natural language prompts. It handles up to 20 URLs per request, converts HTML to clean text, and automatically redirects GitHub blob URLs to raw content. The server blocks private IPs to prevent SSRF attacks and lets you configure timeouts and content limits. You'll need a Gemini API key to run it, and it uses the gemini-2.5-flash model by default. Reach for this when you want Claude to process web content but want the heavy lifting done by Gemini's summarization capabilities rather than feeding raw HTML into context.
claude mcp add --transport stdio kylinmountain-web-fetch-mcp uvx web-fetch-mcp