Which AI agents support Markdown content negotiation?
Skip to content<br>Search the site No matches.
Status<br>AI agent support matrix.
Which AI agents send Accept: text/markdown (or otherwise<br>advertise a Markdown preference) when their built-in browse or fetch<br>tools hit a URL.
Copy page link<br>Open page actions<br>Share<br>Share on X<br>(opens in new tab)<br>Share on LinkedIn<br>(opens in new tab)<br>Share on Bluesky<br>(opens in new tab)<br>Share on Reddit<br>(opens in new tab)<br>Share on Mastodon<br>(opens in new tab)<br>Other…
Copy link
Copy Markdown link
View as Markdown
Open in ChatGPT<br>(opens in new tab)<br>Open in Claude<br>(opens in new tab)<br>Copy Claude Code prompt
Copy Codex prompt
Copy Gemini CLI prompt
Verified an agent's behavior — or spotted something wrong?<br>Send feedback with evidence (server log,<br>curl capture, or vendor statement).
Agent<br>Status<br>Mechanism<br>Verified
Claude Code Anthropic<br>Supports<br>Accept: text/markdown, text/html, */*<br>RFC 7763<br>RFC 9110
2025-11-13
Cursor Anysphere<br>Supports<br>Accept: text/markdown, text/plain;q=0.9, */*;q=0.8<br>RFC 7763<br>RFC 9110
2026-04-18
OpenClaw OpenClaw<br>Supports<br>Accept: text/markdown, text/html;q=0.9, */*;q=0.1<br>RFC 7763<br>RFC 9110
2026-05-04
OpenCode SST<br>Supports<br>Accept: text/markdown;q=1.0, text/x-markdown;q=0.9, text/plain;q=0.8, text/html;q=0.7, */*;q=0.1<br>RFC 7763<br>RFC 9110
2026-05-04
Codex CLI OpenAI<br>Partial<br>Follows<br>RFC 7763<br>RFC 8288
2026-04-18<br>Note Fetches the canonical URL as HTML first (no text/markdown Accept preference), then parses the response for in the document and makes a second request for the .md sibling.
Aider Aider<br>No<br>Fetches only HTML<br>2026-05-09
ChatGPT (browse) OpenAI<br>No<br>Fetches only HTML<br>2026-04-18
Claude.ai (web app) Anthropic<br>No<br>Fetches only HTML<br>2026-04-18
Cline Cline<br>No<br>Fetches only HTML<br>2026-05-09
Copilot Chat (VS Code) GitHub / Microsoft<br>No<br>Fetches only HTML<br>2026-04-18
Copilot CLI GitHub / Microsoft<br>No<br>Fetches only HTML<br>2026-04-18
Devin Cognition<br>No<br>Fetches only HTML<br>2026-05-09
Gemini (web app) Google<br>No<br>Fetches only HTML<br>2026-04-18
Gemini CLI Google<br>No<br>Fetches only HTML<br>2026-04-18
Grok xAI<br>No<br>Fetches only HTML<br>2026-04-18
Microsoft Copilot Microsoft<br>No<br>Fetches only HTML<br>2026-05-09
Perplexity Perplexity<br>No<br>Fetches only HTML<br>2026-04-18
v0 Vercel<br>No<br>Fetches only HTML<br>2026-05-09
Windsurf Cognition<br>No<br>Fetches only HTML<br>2026-05-09
Zed Zed Industries<br>No<br>Fetches only HTML<br>2026-05-09
Last updated 2026-05-09.
Verify an AI agent yourself
We test the matrix ourselves, but AI agent behavior shifts across<br>versions, plans, and newly-added tools. You can corroborate any row<br>— or catch a regression — by triggering an agent against<br>your own server and reading the request from your access logs.
1. Capture the Accept header in your logs<br>Default log formats usually drop it. Add it once:<br>NginxApacheCaddy<br>log_format with_accept '$remote_addr - $remote_user [$time_local] '<br>'"$request" $status $body_bytes_sent '<br>'"$http_referer" "$http_user_agent" '<br>'accept="$http_accept"';<br>access_log /var/log/nginx/access.log with_accept;<br>LogFormat "%h %l %u %t \"%r\" %>s %b \"%{Referer}i\" \"%{User-Agent}i\" accept=\"%{Accept}i\"" with_accept<br>example.com {<br>log {<br>output file /var/log/caddy/access.log<br>format json<br># JSON access logs include request.headers.Accept by default.
2. Trigger a known AI agent against a specific URL
Pick a URL on your site — ideally a unique or freshly-published<br>one so the request isn't masked by background traffic. Then ask the<br>agent to fetch or summarize it:
ChatGPT — "Summarize https://yoursite.com/article-xyz" (with the browse tool enabled)<br>Claude — "What does https://yoursite.com/article-xyz say?" (requires web_fetch / web_search)<br>Perplexity — paste the URL directly into a query<br>…and so on for whichever agent you're testing.<br>3. Report what you saw
Grep for the URL in your access log and send the line(s) to<br>feedback, along with the agent you used<br>and the prompt you sent. The accept= field tells us<br>whether text/markdown was advertised.