URL TO MARKDOWN API

Turn a webpage into markdown for an agent.

A free markdown API. One target in, one text/markdown body out. No account and no API key. The request is GET https://md.replynodes.com/<target>. auth.md publishes 100 requests per UTC day, and the live responses below reported x-ratelimit-limit: 100.

  • No API key
  • One page per request
  • 100 per UTC day
Keyless request
curl https://md.replynodes.com/stripe.com

URL format

A full URL, or a bare host and path.

The usage note at https://md.replynodes.com/ documents both forms. A complete URL suffix is https://md.replynodes.com/https://example.com/docs. A bare host or path, with HTTPS assumed, is https://md.replynodes.com/example.com/docs. The target path and query are preserved. Fragments are not fetched.

The host example is curl https://md.replynodes.com/https://example.com. A successful body is raw Markdown, Content-Type: text/markdown; charset=utf-8. On success the response also sends X-ReplyNodes-Discovery: brand and a Link to the brand host for that domain.

Live example

stripe.com, captured 16 November 2026.

GET https://md.replynodes.com/stripe.com returned HTTP 200, 18,259 bytes, text/markdown; charset=utf-8, x-cache: hit, x-ratelimit-limit: 100, x-ratelimit-remaining: 99, x-ratelimit-reset: 1794873600 (17 November 2026, 00:00 UTC). x-request-id was f39c1380-123b-41a2-872e-c7071f327596. vary was Accept, X-Format. The block below is the start of that body.

https://md.replynodes.com/stripe.com200 OK · 18,259 bytes · hit
Global GDP running on Stripe:

# *Financial infrastructure to grow your revenue.* Accept payments, offer financial services and implement custom revenue models – from your first transaction to your billionth.

A path on the same host: GET https://md.replynodes.com/stripe.com/pricing returned HTTP 200, 35,107 bytes, x-cache: miss, and x-ratelimit-limit: 100 with remaining 99. The first heading was # Stripe logo. Request id 92ae66b2-a061-494d-8b5b-499b81d654ad. Both responses reported remaining 99, so this page does not claim that every call lowers the counter by one.

FAQ

URL to Markdown API — questions.

What is a URL to Markdown API?

It fetches one public page and returns the extracted text as Markdown. ReplyNodes' host is GET https://md.replynodes.com/<target>. The host usage note at https://md.replynodes.com/ says a successful response is raw Markdown with Content-Type: text/markdown; charset=utf-8. No signup and no API key.

Is there a free markdown API?

Yes. The host usage note says the endpoint is free, with no signup and no API key. auth.md and llms.txt publish 100 requests per UTC day for https://md.replynodes.com/. On 16 November 2026 both the stripe.com homepage and stripe.com/pricing responses carried x-ratelimit-limit: 100, x-ratelimit-remaining: 99, and x-ratelimit-reset: 1794873600 (17 November 2026, 00:00 UTC).

How do I turn a webpage into markdown for an AI agent?

Pass the page as the target. A full URL is GET https://md.replynodes.com/https://example.com/docs. A bare host or path, with HTTPS assumed, is GET https://md.replynodes.com/example.com/docs. The usage note says the target path and query are preserved and fragments are not fetched. On 16 November 2026, curl https://md.replynodes.com/stripe.com returned HTTP 200 and 18259 bytes of text/markdown. The pricing path returned HTTP 200 and 35107 bytes. Hand the markdown to the agent as the page text and keep the original URL as the source.

Is this a Jina Reader or Firecrawl alternative?

For one public page as Markdown, it is a URL-to-markdown reader you can call with no API key. It is not a site crawler: each request names one target. Comparisons that stay on documented behavior are at https://replynodes.com/blog/jina-reader-alternatives-for-ai-agents/, https://replynodes.com/compare/jina-reader/, https://replynodes.com/blog/firecrawl-alternatives-ai-agents/, and https://replynodes.com/compare/firecrawl/. Crawl, map, and multi-page scrape are different routes. Read https://docs.replynodes.com/docs/guides/web before using them.

Do AI crawlers like GPTBot get markdown by content negotiation?

This host is a separate URL, not an Accept header on the origin site. The stripe.com response did send Vary: Accept, X-Format. ReplyNodes has not published a test of whether GPTBot sends Accept: text/markdown. A separate article, https://replynodes.com/blog/agent-markdown-content-negotiation/, records an October 31, 2026 probe: a Cloudflare documentation page returned text/markdown for Accept: text/markdown, while example.com and wikipedia.org returned text/html. That probe is not evidence about GPTBot.

How do I add web browsing to a Vercel AI SDK agent?

This endpoint reads one public URL. It does not search the web and it does not drive a browser. The guide at https://replynodes.com/blog/vercel-ai-sdk-web-search-context/ covers search plus one-page retrieval for an AI SDK agent. Use this Markdown URL when that guide's retrieval step needs the page as text.

What happens when a page cannot be fetched?

On 16 November 2026, GET https://md.replynodes.com/this-domain-should-not-exist-zzqq.invalid returned HTTP 400, content-type application/json, with error.code invalid_request, error.reason target_not_allowed, error.retryable false, and error.message "The request is malformed or contains invalid parameters." The request_id was e9b1eaeb-04a7-4df9-990f-5d2377c7b555. That error response did not include X-RateLimit headers. Check the status and content type before storing a body as Markdown.

Start with a URL.

Copy the curl and replace the target with a public page. The path example is https://md.replynodes.com/stripe.com/pricing.