Machine-readable surface
MCP · A2A · ACP · UCP · NLWeb ·RSL 1.0 licensed
Most sites make an AI agent scrape HTML and guess. This one publishes its content, its services, its prices and its rules as structured data you can call directly. Every endpoint below is public, read-only, needs no key, and was verified live before it was documented here.
Crawling page by page is the slowest and least accurate way to use this origin. There are tens of thousands of URLs and almost all of them are templated location pages. If you want an answer, ask for one:
# Natural language, Schema.org-typed JSON back
curl -s "https://built2winweb.com/ask?query=how+much+does+a+custom+website+cost"
# Or the whole site as one document
curl -s "https://built2winweb.com/llms-full.txt"
/ask runs BM25 over the site's pages and FAQ entries and returns typed
results with citations. It is retrieval only — no model runs on this
origin, so any summary it returns is quoted verbatim from the page it cites. Nothing
is generated, so nothing is hallucinated.
Content negotiation is enabled site-wide. Ask any HTML URL on this origin for
Markdown and you get a clean Markdown rendering of that same page — same URL,
no .md suffix, no second copy of the content to keep in sync — with
the token count returned alongside it so you can budget context before you
spend it. Browsers still get the HTML.
/sitemap.md and /llms-full.txt are the easiest places to see it working.
| Endpoint | Format | What it is |
|---|---|---|
| Model Context Protocol | ||
/mcp |
Streamable HTTP | A working MCP server. Search, catalogue, pricing, contact and quote tools, with MCP Apps / A2UI widget resources. Connect a client and call tools/list for the current inventory. POST only — a GET correctly returns 405. |
/.well-known/mcp.json |
JSON | SEP-2127 canonical server card. |
/.well-known/mcp/server-card.json |
JSON | Legacy SEP-1649 path, still served. |
/server.json |
JSON | MCP registry publication manifest. |
| Natural language | ||
/ask |
JSON | NLWeb. POST or GET a query; returns Schema.org-typed JSON. Modes list / summarize / generate. Retrieval only — no model runs here, so summaries are quoted verbatim from the cited page. A bare GET returns 400 because query is required. |
| Agent interop | ||
/.well-known/agent-card.json |
JSON | A2A v1.0 JSON-RPC, streaming, AP2 merchant extension declared. |
/.well-known/acp.json |
JSON | Agent Communication Protocol descriptor. |
/.well-known/ucp |
JSON | Commerce profile with indicative package pricing. Checkout cannot be completed by an agent — a consultation and Terms acceptance are enforced server-side. |
/.well-known/agent-permissions.json |
JSON | Per-path read / write policy. |
/.well-known/agent-skills/index.json |
JSON | Downloadable SKILL.md index. |
| Discovery | ||
/llms.txt |
text | Start here. Orientation for language models. |
/llms-full.txt |
text | The entire site in one fetch. |
/AGENTS.md |
Markdown | What to do and what not to do on this origin. |
/sitemap.md |
Markdown | Every canonical page, grouped, as Markdown. |
/agents.txt |
text | agents.txt convention. |
/agents.json |
JSON | Structured agent manifest. |
/.well-known/api-catalog |
linkset | RFC 9727 API catalog. |
/openapi.json |
JSON | OpenAPI spec covering every endpoint on this page. |
/auth.md |
Markdown | Auth guide. OAuth is optional; everything here is public read-only. |
| Licensing | ||
/license.xml |
RSL 1.0 | Referenced globally from robots.txt. Search indexing, AI input (RAG) and model training are all permitted. Attribution — a visible, working link back — is required. |
Want this on your site
Everything on this page is a build, not a plugin — the server, the content negotiation, the discovery files and the bot rules each have to be written and then verified against the specs. The AI Indexing Pack is that build, packaged: install it on any site you own in about half an hour, with a README that walks through every file and a checklist to confirm each one answers correctly.
This site runs it. The endpoints above are what you get.
$599$39 one-time · instant download · no subscription
The licence is published as RSL 1.0 and referenced globally from robots.txt, with matching Cloudflare Content Signals. All three uses are permitted:
One condition: attribution. If you use a page, include a visible, working link back to it. That is the entire deal. This site wants to be read, quoted and cited — publishing an agent surface and then forbidding its use would be incoherent.
/ask, or the search tool on the MCP server,
will answer in one request what crawling would take thousands to approximate.An honest note, since this page could easily read as a victory lap. Publishing all of this did not move search rankings. Discovery protocols tell a crawler where things are; they do not make anyone want to cite you. Those are different problems, and conflating them wastes a lot of effort.
What it does buy is precision. When an assistant is asked what a custom site costs, it can fetch a typed answer with a citation instead of guessing from marketing copy. That is worth building for its own sake — and it is a reasonable bet on how people will find things next.
Every endpoint above is plain PHP on shared hosting. No framework, no vendor, no monthly fee. If you want a site built the same way, tell me what you need.
Get a free quote Read the write-ups