agenticweb.wiki

Markdown for Agents

Markdown for Agents is an agentic-web standard (HTTP header) at step 4, Read, governed by Cloudflare; its own status is "Vendor feature, available to Pro, Business and Enterprise plans and SSL for SaaS customers at no cost; not a standards-track document" (living). It answers: Can an agent ask a site for a Markdown version of an HTML page, instead of parsing the HTML itself? Adoption is observed by Cloudflare's own developer docs and blog answer Accept: text/markdown with text/markdown (our check, 2026-10-20); Cloudflare publishes no count of enabled zones, last verified 2026-10-20.

Read Markdown for Agents is a Cloudflare network feature: when an agent requests a page with Accept: text/markdown, an enabled zone fetches the HTML from the origin, converts it to Markdown at the edge and returns it as text/markdown with Vary: Accept, so caches keep the HTML and Markdown variants apart [1]. The output has a fixed layout: optional YAML frontmatter built from the page's title, description and og:image meta tags, the body with navigation, scripts and styles stripped, and any JSON-LD kept in a fenced block at the end [1]. Only HTML is converted, up to 6 MiB [1].

Overview

It is a vendor feature, not a standard: any server can do the same content negotiation, and Vercel's Agent Readability Spec lists Accept: text/markdown as one of its checks. Cloudflare's version is free on Pro, Business and Enterprise plans and for SSL for SaaS, and is switched on in AI Crawl Control, per path or host by a Configuration Rule, or through the content_converter zone setting [1]. Cloudflare launched it on 2026-02-12 and serves its own docs and blog this way [2][3].

Two details matter. First, if the origin sends no content-signal header, Cloudflare adds Content-Signal: ai-train=yes, search=yes, ai-input=yes to the converted response [1][3], so a site that turns the feature on without setting its own header advertises consent to every use the Content Signals Policy names. Second, the launch post describes an x-markdown-tokens header with an estimated token count [2], but the current docs do not mention it and our check on 2026-10-20 did not receive it [3].

Discovery and probe

HTTP content negotiation: the client sends Accept: text/markdown and an enabled zone answers with Content-Type: text/markdown; charset=utf-8 and Vary: Accept. There is no manifest or well-known file.

GET {origin}/ (request header Accept: text/markdown)
parse:     Content-Type is text/markdown; a Content-Signal response header is present (Cloudflare adds ai-train=yes, search=yes, ai-input=yes when the origin sets none); an x-markdown-tokens header is described in the launch post but was not returned in our 2026-10-20 check
on absent: An HTML response means the zone has not turned the feature on, or is not on Cloudflare; the site may still offer Markdown by other means, such as .md mirrors or llms.txt

Not probed live: no-fixed-path. Needs an Accept: text/markdown request header on any page; the probe rule format cannot set request headers.

History

Instances

No instances recorded yet.

See also

References

  1. Cloudflare's docs page (last updated 2026-10-09) says a client that sends Accept: text/markdown to an enabled zone gets the HTML converted at the edge, with Content-Type: text/markdown; charset=utf-8 and Vary: Accept. Output is optional YAML frontmatter from meta tags, the body as Markdown, and any JSON-LD in a fenced json block. Only HTML is converted, up to 6 MiB. When the origin sends no content-signal header, Cloudflare adds Content-Signal: ai-train=yes, search=yes, ai-input=yes. It is enabled in AI Crawl Control, by a Configuration Rule, or via PATCH /zones/{zone_tag}/settings/content_converter, and is free on Pro, Business, Enterprise and SSL for SaaS. — https://developers.cloudflare.com/fundamentals/reference/markdown-for-agents/ (2026-10-09) VERIFIED
  2. The launch post (datePublished 2026-02-12) introduces the feature and says converted responses carry an x-markdown-tokens header with the estimated token count of the Markdown document. — https://blog.cloudflare.com/markdown-for-agents/ (2026-02-12) VERIFIED
  3. Our own request on 2026-10-20 to this post and to the docs page with Accept: text/markdown returned HTTP 200, content-type: text/markdown; charset=utf-8 and content-signal: ai-train=yes, search=yes, ai-input=yes, and no x-markdown-tokens header. — https://blog.cloudflare.com/markdown-for-agents/ (2026-10-20) VERIFIED

Disputed: The token-count header promised at launch is not in the current docs and was not returned in our check, so agents should not rely on it. The feature also attaches a permissive Content-Signal by default, which a site owner may not expect from a format-conversion switch.

JSON · Markdown