Free “Markdown for Agents” (Accept: text/markdown) & Content-Signal Checker

Test in 1 click whether your website or Cloudflare edge serves clean Markdown when AI agents send Accept: text/markdown, measure exact token savings vs. raw HTML, and generate Content-Signal policy headers.

Live HTTP Content Negotiation Result for vercel.com

Full AI Authority Report (85/100) →
✓ Accept: text/markdown Active ✓ Vary: Accept Cache Header Set ○ No Content-Signal Header ✓ /llms.txt Fallback Active
Fetch Mode HTTP Header Sent Payload Size AI Tokens Token Savings
Standard Browser HTML Accept: text/html 586.1 KB 150,049 Baseline
Markdown for Agents Accept: text/markdown 2.2 KB 567 -99.6%
Static /llms.txt Summary GET /llms.txt 14.5 KB 3,722 -97.5%
Live Preview of Accept: text/markdown Response from vercel.com
# Agentic Infrastructure

For coding agents to ship apps and agents automated by agents.

- For coding agents to deploy in their native language, with Vercel's API, CLI, MCP, and Skills.

- To ship apps and agents in Sandboxed VMs, with durable backends, powered by 400+ models.

- Automated by agents who autonomously investigate errors, plan fixes, and open PRs.

[Deploy now](/new)

[Talk to sales](/contact/sales)

## Build agents on infrastructu

1-Click Content-Signal & Markdown Negotiation Config Builder

Cloudflare · Apache .htaccess · Nginx

Frequently Asked Questions

Technical Reference

What is HTTP Accept: text/markdown ("Markdown for Agents")?

Markdown for Agents uses standard HTTP content negotiation: when an AI coding agent or crawler requests a URL with the header Accept: text/markdown, an edge network (such as Cloudflare) or origin server returns clean Markdown instead of bloated HTML—typically cutting token payloads by 80% to 99%.

What does the Content-Signal HTTP header control?

Content-Signal lets publishers declare machine-readable AI permissions across three directives: search (search result indexing), ai-input (real-time RAG/answer grounding), and ai-train (foundation model pre-training)—for example: Content-Signal: search=yes, ai-input=yes, ai-train=no.