read-url

name	read-url
description	Fetch web content as clean Markdown using Exa, Jina Reader, and WebFetch. Choose the right tool for each scenario.

/read-url — Fetch Web Content as Clean Markdown

Turn any URL into readable Markdown. Three tools, pick the right one.

Tool Selection

Exa — mcp__Exa__crawling_exa

Primary tool. Clean Markdown extraction with configurable maxCharacters.

mcp__Exa__crawling_exa
  urls: ["https://example.com/article"]
  maxCharacters: 6000  # default 3000, adjust as needed

Also use mcp__Exa__web_search_exa for search-first-then-fetch workflows.

Jina Reader — r.jina.ai

Fallback when Exa can't reach the content (paywalled, JS-heavy, blocked).

https://r.jina.ai/YOUR_URL

Use with WebFetch:

WebFetch
  url: "https://r.jina.ai/https://example.com/article"
  prompt: "extract the main content..."

Jina Search — s.jina.ai

For searching when you need to find pages first, then read them.

https://s.jina.ai/YOUR_SEARCH_QUERY

WebFetch

Built-in fallback. Use when Exa and Jina both fail, or for quick single-page fetches.

WebFetch
  url: "https://example.com"
  prompt: "extract key information about..."

Decision Flow

Can Exa reach it?
  YES → mcp__Exa__crawling_exa
  NO  → Is it JS-heavy / paywalled / blocked?
          YES → https://r.jina.ai/URL via WebFetch
          NO  → WebFetch directly

Need to search first?
  → mcp__Exa__web_search_exa (primary)
  → https://s.jina.ai/QUERY (fallback)

When to Use

• Reading documentation, articles, or references at a URL
• Fetching GitHub READMEs, issues, PRs (prefer gh CLI for GitHub)
• Extracting content from paywalled or JS-heavy pages
• Searching the web for information, then deep-reading results

When NOT to Use

• GitHub API operations → use gh CLI
• Local files → use Read tool
• Interactive auth-required pages → use specialized MCP tools

References

• Jina Reader — r.jina.ai for reading, s.jina.ai for searching
• Exa MCP tools — mcp__Exa__crawling_exa, mcp__Exa__web_search_exa