---
title: "AI-Ready Microsites and PDFs, Out of the Box | RELAYTO"
description: "Every public RELAYTO workspace site is AI-ready by default: Markdown of every page and PDF, an MCP server, llms.txt and AI crawler rules, kept current."
canonical: "https://relayto.com/ai-ready"
---

# AI-ready microsites and PDFs, with nothing to set up

> Every public RELAYTO workspace site is AI-ready by default: Markdown of every page and PDF, an MCP server, llms.txt and AI crawler rules, kept current.

- Web page: https://relayto.com/ai-ready
- Index of relayto.com for agents: https://relayto.com/llms.txt

**4.7** from 470+ reviews on G2 **4.8** on Capterra

RELAYTO makes your microsites and PDFs AI-agent friendly: a clean Markdown copy of every page and document, plus llms.txt, an MCP server and WebMCP tools, kept current as you publish.

[Request a demo](https://relayto.com/book-demo)

## Rated by independent audits

**[ora.ai](https://ora.ai/score/benefits.relayto.com)**
: Grade A Agent-Ready

**[is-agentic.com](https://is-agentic.com/scan/benefits.relayto.com)**
: 98/100 Strong technical baseline

**Cloudflare**
: Agent-Integrated Agent Readiness Scanner

Calder Group Annual Report at investors.yourcompany.com, two ways. For people: the report open at its cover in RELAYTO's reader, “Energy for a lower-carbon decade”, with its navigation beside it. For AI agents: the same report as Markdown, starting with its title, description, address and date, who published it, “PDF document, 10 pages”, then page 1's text.

Used by listed companies with a **$2 trillion** combined market cap.

- ![Franco-Nevada](https://relayto.com/_astro/logo-franco-nevada.DaecV64H_1W57eI.webp)
- ![Omnicom Group](https://relayto.com/_astro/logo-omnicom.DErW2nMs_Z1EyOM8.webp)
- ![Salesforce](https://relayto.com/_astro/logo-salesforce.Dn5UI1aO_24lE8M.webp)
- ![Westinghouse](https://relayto.com/_astro/logo-westinghouse.BmPRQZQh_Z2ff40g.webp)
- ![LHD Benefits](https://relayto.com/_astro/logo-lhd.CdehYuhD_Z1l4cLf.webp)

## Every PDF and page, in clean Markdown

RELAYTO writes a Markdown copy of every public page, hub and PDF: plain, structured text that AI reads directly. PDFs come page by page, their tables as real Markdown tables.

Document

Annual report (PDF) Investor web page

View as

Person AI agent

Page 5 of the Calder Group Annual Report in RELAYTO's reader: “Earnings growth ahead of guidance.”, a table of results in £ millions for this year and last year (underlying EBITDA 1,310, up 12%) and a line chart of revenue by quarter.

The same page in the report's Markdown copy: under “Page 5”, its heading, the results table as a Markdown table with every row and column (underlying EBITDA 1,310, 1,170, +12%), and the picture on the page described in a sentence.

Calder Group's investor web page “Share information” at investors.yourcompany.com: a header photograph of a transmission line at dusk, then a table of results at a glance, this year against last (revenue £4,820m, underlying EBITDA £1,310m, dividend 41.6p).

The web page's Markdown copy: the photograph as its alt text, then the results table as a Markdown table with every row and column (revenue 4,820, underlying EBITDA 1,310, dividend 41.6).

About 8×

smaller than the page an AI crawler would otherwise read (7.5× to 9.2× on three RELAYTO pages), so an AI takes in more of each document at once.

- **Every page, hub and PDF**

  Microsites, hubs and documents, each at its address plus .md.
- **Page numbers to cite**

  Each PDF page under its own heading, so an answer can name it.
- **Web pages keep their tables**

  Headings, lists, tables and each image’s alt text, as on the page.
- **Layout-aware PDF parsing**

  Marker, a layout-aware PDF parser, turns tables into Markdown tables and describes images.

## Agents can search your site and quote it

Claude, ChatGPT and other AI tools can search your investor documents, read the document they need and quote it with a link back. Reading your site needs no sign-in.

An AI assistant is asked “What was Calder Group’s underlying EBITDA last year?” It calls search_content and then get_content on investors.yourcompany.com/mcp: the search finds the Calder Group Annual Report, and reading it returns the results table on page 5. It answers: “Underlying EBITDA was £1,310m, up 12% from £1,170m the year before.”, citing Calder Group Annual Report, page 5 at investors.yourcompany.com/e/calder-group-annual-report-b2n6x4kr8tf9e.

**MCP server `/mcp`**
: AI tools connect to your site, search it and read any page.

**MCP Registry `com.relayto/site-content`**
: One official MCP Registry entry covers your site’s server, at your address.

**WebMCP `workspace-tools-v2.js`**
: Agents working in a browser use your site’s own search.

**Agent Skills `/.well-known/agent-skills`**
: A ready instruction card: how to find, read and cite you.

**Site API and OpenAPI `/api/site/v1`**
: Your published content as data, with no API key.

**NLWeb `/ask`**
: A question in plain words returns the matching pages.

**CLI and SDK `npx relayto`**
: Developers search your site from a terminal or their code.

## Found by AI search, and understood

Each site publishes the files AI crawlers and agents look for first, and gives crawlers that cannot run scripts the finished page.

Calder Group Investor Centre at investors.yourcompany.com, a RELAYTO hub with its sustainability and results hubs and its reports. Over it, the site's llms.txt: Calder Group Investor Relations, a one-line summary, the tools for AI agents (the MCP content server, authentication for agents, the agent skill), then its hubs and documents, each linked to a Markdown copy.

- **llms.txt**

  A guide for AI: who you are, every hub and document, the tools.
- **Markdown at the same address**

  Ask any page for text and it answers in Markdown.
- **Full pages for crawlers**

  Search and AI crawlers get each page fully rendered, text included.
- **Catalogues and structured data**

  Standard files listing every page, what each one is, and your AI tools.

## Search and answers on, training off by default

Your site lets AI companies use your content for search and answers. Model training is off by default, and you decide whether to switch it on. Agents see only what an anonymous visitor could open.

The robots.txt of investors.yourcompany.com: for every crawler, the Content Signal search=yes, ai-input=yes, ai-train=no; a group for AI search and for assistants fetching a page for a user (OAI-SearchBot, ChatGPT-User, Claude-SearchBot, Claude-User, PerplexityBot and more) with the same signal; then the sitemap. Beside it, the policy in plain words: search yes, AI answers yes, model training off by default.

- **Content Signals**

  One robots.txt line tells every crawler: search yes, AI answers yes, training off.
- **Training off by default**

  Model training stays off unless you switch it on for your site.
- **Rights reserved**

  A machine-readable notice, made for EU copyright rules, reserves text and data mining.
- **Closed pages stay closed**

  Private, password, sign-in and gated pages stay closed to agents.

[How Content Security controls access](https://relayto.com/product/content-security)

## What AI reads changes when you publish

Publish, update or take down a document, and its Markdown, llms.txt, sitemap and search follow within minutes. New standards reach every workspace as RELAYTO adds them.

A report, Calder Group Annual Report, is published. RELAYTO then updates, on its own: calder-group-annual-report-b2n6x4kr8tf9e.md (markdown copy), llms.txt (listed), sitemap.xml (listed), /mcp (in agent search), /api/site/v1 (in the site api), /schema-feed.jsonl (in the schema.org catalogue).

- **Publish**

  A new report is listed and searchable by agents.
- **Update**

  Replace the PDF and its text is read again.
- **Take down**

  Unpublished or private content leaves every file and the search.
- **New standards**

  RELAYTO adds them to every workspace at once.

## Four layers RELAYTO adds to every site

People see your investor site as it is. Under it, every public workspace site gets four layers that AI search and agents use, all on by default.

1. **Readable**

   Every page and PDF as clean text an AI can read and cite.

   - `<page>.md`
   - `Accept: text/markdown`
   - `## Page N`
   - `Marker`
   - `Prerender`
2. **Findable**

   A map of everything you publish, in the files AI search looks for.

   - `llms.txt`
   - `sitemap.xml`
   - `schema.org feed`
   - `Link headers`
   - `ARD`
   - `RFC 9727 api-catalog`
3. **Usable by agents**

   Tools that let AI search your site, read it and quote it.

   - `MCP server`
   - `MCP Registry`
   - `WebMCP`
   - `Agent Skills`
   - `OpenAPI 3.1`
   - `NLWeb /ask`
4. **Your rules**

   Search and AI answers allowed, model training off, closed pages closed.

   - `robots.txt`
   - `Content Signals`
   - `TDMRep`
   - `auth.md`

All four follow what you publish, within minutes. `max-age=300`

[Developer docs for your site’s API and agents](https://relayto.com/developers)

## Common questions

Set-up, training, domains, privacy and PDFs.

**Do we need to do anything?**

No. Every public RELAYTO workspace site gets all of this, on a relayto.com subdomain or your own domain. There is nothing to install, switch on or maintain. Publish as you do today, and the Markdown copies, llms.txt, sitemap and agent tools follow each change on their own.

**Will AI companies train on our content?**

Not unless you want them to. Training is off by default: your site’s robots.txt allows search and AI answers and refuses model training, and a TDMRep file reserves text and data mining rights. If you want your content used to train AI models, you can switch training on for your site. Crawlers that follow these standards respect your choice; no website can force one that ignores them.

**Does this work on our own domain?**

Yes. A site on your own domain, such as investors.yourcompany.com, gets the same Markdown copies, llms.txt, crawler rules and agent tools as one on a relayto.com subdomain, all at your own address. A customer’s investor relations site on its own domain runs these agent features today.

**What about private documents?**

Agents see only what an anonymous visitor could open. Private, password-protected, sign-in and email-gated pages stay closed, and a page behind a disclaimer is not reproduced. Unlisted pages stay out of llms.txt, the sitemap and agent search; anyone with the link can still read them. One privacy rule covers the Markdown copies, the MCP server, WebMCP and the API.

**Which AI tools does this work with?**

Assistants that open a page for a user, such as ChatGPT, Claude, Perplexity and Mistral, get the Markdown copy automatically, and any tool can ask for it. Claude, ChatGPT and other AI tools that support MCP can connect to your site and search it, and agents working in a browser can use its WebMCP tools.

**How well does AI read our PDFs?**

Every PDF is readable page by page, so an assistant can say which page a figure came from. RELAYTO parses it with Marker, a layout-aware PDF parser, which turns tables into real Markdown tables and describes images in words.

**How is this different from adding llms.txt ourselves?**

An llms.txt you write is one file, out of date at your next publish. RELAYTO writes yours from what is published, with a Markdown copy of every page and PDF behind each link, and adds the MCP server, crawler rules and discovery files agent-readiness audits check. All of it follows your content.

## See a site the way AI agents do

We’ll open a live RELAYTO site and show you what ChatGPT, Claude and other agents read from it.

[Request a demo](https://relayto.com/book-demo)
