# Agent Ready HTML: notes for agents

Agent Ready HTML is a publication about what an agent receives when it fetches a page without running JavaScript. It covers the initial HTML, real status codes, llms.txt, markdown on request, JSON-LD and OpenAPI links. It is a set of static pages and downloads with a read-only API, MCP server and ask endpoint over the same data, all public.

## When to read it

Read this site when you need to know whether a page or a platform is readable by a plain-fetch client, what an llms.txt should contain, or how common a read-layer practice is among the top 10,000 domains.

## Which page answers which question

- What does a given URL give a plain fetch? The viewer at /viewer/?url=<url>. It makes four GET requests and flags the H1, the first 200 words, llms.txt, markdown, the status of a missing path and JSON-LD.
- Does a given site have an llms.txt, and how does it grade? /llms-txt-index/<domain>/, with the file, the parse and six structure checks.
- How common is a practice? /data/ and /llms-txt-index/2026-11-04/, with CSV and JSON downloads.
- How does a docs platform or site builder score? /platforms/<id>/, with the evidence and read date for each key.
- What is a term? /glossary/<slug>/. Short answers: /faq/. Procedures: /guides/.

## Formats

Every page has a markdown twin: send Accept: text/markdown, or add /index.md to the path. The root /llms.txt lists the site, and section indexes such as /llms-txt-index/llms.txt list the pages below them. /llms-full.txt holds every page in one file. /feed.xml and /sitemap.xml list what changed.

## Query it as data

- MCP server: `POST /mcp` (Streamable HTTP, JSON-RPC 2.0, no credential). Tools: `search_llms_index`, `get_llms_entry`, `grade_distribution`, `get_crawl_measure`, `list_platforms`. Card at `/mcp/server-card`, manifest at `/.well-known/mcp.json`.
- REST API: `/api` is the index, one GET endpoint per tool under `/api/v1/`, described by `/openapi.json` and linked from `/.well-known/api-catalog`.
- Ask: `/ask?query=...` (NLWeb 0.55, GET or POST, `streaming=true` for events) returns the matching rows and pages as schema.org objects with a source URL each.
- Skills: `/.well-known/agent-skills/index.json` lists two skills that show the calls.
- Everything is read-only and documented at `/developers`; the rate limit is 120 requests a minute per address.

## Sourcing

Every fact carries a source URL and the date it was read. Statements a vendor makes about itself are labelled as vendor claims. Platform scores come from a dated test build where the platform builds locally without an account (seven platforms), and from dated documentation otherwise; each platform page says which.

## Limits

The site is a publication. Its API, MCP server and ask endpoint read the same data files the pages show; nothing can be written through them and no account exists. The viewer is a web page that accepts a URL; it is limited to ten runs per address per minute.