For AI agents

This site speaks agent

An MCP server, Markdown for every page, WebMCP tools and discovery documents, so AI assistants can find, read and cite 63Digital accurately.

This site is built to be read, searched and cited by AI agents as easily as by people. Everything below is public, read-only and needs no sign-up. If you are an agent, start with the MCP server or request any page with Accept: text/markdown.

MCP server

A public Model Context Protocol server at https://63digital.com/mcp (Streamable HTTP, no authentication). Add it to Claude, ChatGPT, Cursor or any MCP client as a remote server, then ask about our services, past work or guides.

  • list_services: our services, with a one-line summary each.
  • list_case_studies: client case studies and their outcomes.
  • list_platform_guides: the platforms we work with and the practical guides written for each.
  • latest_updates: recent posts from Updates.
  • search_site: full-text search across every public page.
  • get_page: any page as Markdown.
  • contact_details: how to start a project.

Server card: /.well-known/mcp/server-card.json.

Ask 63Digital: an A2A agent

An Agent2Agent agent at https://63digital.com/a2a (JSON-RPC, A2A 1.0 and 0.3) answers questions about our services, platforms and past work in plain language, using the same tools as the MCP server. Agent card: /.well-known/agent-card.json.

Sending an enquiry: auth.md

Reading is open, but sending us an enquiry needs the person's say-so. Agents register with the person's email per auth.md: we email them a sign-in link, they enter the 6-digit code their assistant shows them, and the agent gets a short-lived token for POST /api/enquiries or the MCP submit_enquiry tool. Replies go to the verified address. Metadata: /.well-known/oauth-protected-resource.

Buy us a coffee: MPP and ACP

The one thing for sale is a £3 coffee for the team, settled through Stripe. Agents pay with a Stripe Shared Payment Token, either over MPP (POST /coffee/buy answers 402 with a payment challenge) or the Agentic Commerce Protocol (/.well-known/acp.json, with an access token from auth.md). People can use the coffee page.

Pages as Markdown

Request any page with the header Accept: text/markdown and you get its main content as clean Markdown, with the title, description and canonical URL in front matter and an x-markdown-tokens estimate. Browsers keep getting HTML.

In the browser: WebMCP

In browsers that support WebMCP, every page registers tools an agent can call directly: search the site, read a page as Markdown, open a page, and fill in our contact form with an enquiry. The form is never sent automatically; the person reviews it and presses Send.

Discovery

  • Agent skills: instructions for reading this site and for helping someone start an enquiry.
  • API catalog (RFC 9727) and AI catalog (Agentic Resource Discovery).
  • robots.txt: all crawlers welcome, with Content-Signal: search=yes, ai-input=yes, ai-train=yes.
  • Link headers on the home page point to all of the above.

Want this for your site?

Making a site agent-ready is mostly careful engineering on what you already have: clean Markdown, honest discovery documents and a small MCP server over your real content. We built all of this on our own Umbraco site and can do the same for yours. Book a discovery call.