Using Markdown New

Using markdown.new for Web Content Fetching

Overview

markdown.new is a Cloudflare-powered service that converts any URL to clean Markdown, reducing token usage by approximately 80% compared to raw HTML. When fetching web content, always use markdown.new instead of direct URLs to optimize context window usage.

Basic Usage Pattern

Standard Conversion

To fetch web content, prepend https://markdown.new/ to the target URL:

Target URL: https://example.com/blog/post
markdown.new URL: https://markdown.new/https://example.com/blog/post

Use with WebFetch:

WebFetch("https://markdown.new/https://example.com/blog/post", "summarize this article")

With Images

By default, markdown.new strips images to minimize tokens. To retain images, add retain_images=true:

WebFetch("https://markdown.new/https://example.com/docs?retain_images=true", "analyze the documentation")

When to use this: Documentation includes diagrams, design mockups, visual content, or screenshots that are essential for the task.

JavaScript-Heavy Sites

For single-page applications or JavaScript-heavy sites, use the browser rendering method:

WebFetch("https://markdown.new/https://example.com?method=browser", "extract content from this React app")

The service automatically tries the fastest method first and falls back as needed, but specifying method=browser ensures full page rendering.

Token Efficiency

markdown.new reduces token usage dramatically:

• HTML: <h2 class="section-title" id="about">About Us</h2> → 12-15 tokens
• Markdown: ## About Us → 3 tokens
• Reduction: ~80% fewer tokens

This means 5x more web content fits in the same context window.

When to Use

Always use markdown.new when:

• Fetching blog posts, articles, or documentation
• Scraping web pages for analysis
• Extracting content from websites
• Building knowledge bases or RAG pipelines
• Any task requiring WebFetch

Exception: Skip markdown.new for authenticated sites or APIs that return structured JSON (use direct fetch instead).

Additional Resources

For detailed information, consult:

Reference Files

• references/conversion-methods.md - Detailed explanation of auto/ai/browser conversion pipeline
• references/query-parameters.md - Complete documentation of method and retain_images parameters
• references/response-format.md - What to expect in responses (headers, token counts, metadata)
• references/edge-cases.md - Troubleshooting and when to fallback to direct WebFetch

Working Examples

• examples/fetch-blog-post.md - Example fetching a blog article
• examples/fetch-with-images.md - Example using retain_images for visual content
• examples/fetch-js-heavy-site.md - Example using browser method for SPAs

Quick Reference

Always use markdown.new for web content fetching to optimize token usage and context window efficiency.