Agent Fetch

Claude's built-in fetch truncates or summarizes pages, which is annoying when you actually need the full article. This gives you complete text extraction with structure intact (headings, lists, links) using multiple fallback strategies and browser impersonation. It's fast at 200-700ms and runs locally via npx, no server needed. Works for single URLs, crawling multiple pages, PDFs, and has options for custom selectors and cookie handling. The JSON output mode is clean for parsing. Honest take: this should probably be the default fetch behavior, but until then, reach for this whenever you're reading URLs and need everything on the page.

Install to Claude Code

npx -y skills add teng-lin/agent-fetch --skill agent-fetch --agent claude-code