This site is set up so AI agents can read it directly, without screenshots or scraping. If you use an AI assistant to research local gardeners, compare services or check what we cover, it can ask this site for the answer itself.

There are three ways in. Use whichever one your agent supports.

1. In a browser with a built-in AI agent

Browsers that support WebMCP see this site’s tools on every page automatically. There is nothing to install or connect. Ask your browser’s agent something like “what services does Daylesford Gardener offer?” and it will use the tools below instead of reading the page.

WebMCP is a proposed web standard. It is currently in early preview in Chrome and has to be switched on there first.

2. From a desktop AI app (Claude, Cursor, Cline, Windsurf)

Desktop MCP clients connect through a small bridge that runs on your computer. This works on this page only.

1. Install the bridge. Run this in a terminal, replacing claude with cursor, cline or windsurf if you use one of those:

npx -y @jason.today/webmcp@latest --config claude

2. Get a token. Ask your agent to generate a WebMCP token, or run:

npx @jason.today/webmcp --new

3. Connect. Click the </> agent button in the bottom-right corner of this page, paste the token and press Connect. A green dot on the button means your agent is connected and can use the tools below.

3. With no setup at all

• Markdown pages. Request any page with the header Accept: text/markdown and you get clean Markdown instead of HTML. Adding ?format=md to a page address does the same in a normal browser.
• Content API. The same tools are available as a read-only API at /wp-json/scos/v1, described in OpenAPI.
• Discovery files. Agents can find everything above without being told, through the API catalog, the MCP server card, the agent skills index and the ARD catalog.

Available tools

All four tools are read-only. They return published pages only.

get_site_info

Who we are, what we do, the area we cover and our services. Agents should call this first.

search_content

Search service pages, projects and articles by keyword, for example “hedge trimming” or “Airbnb”. It can be narrowed by content type or by what the visitor is trying to do.

list_content

List everything published, with the same optional filters. This is useful for seeing every service before looking at one.

get_content

The full text of one page or article, found by its slug from a search or list.

Asking for a quote

An agent can help you ask for a quote, but it never sends one for you. In a browser with a built-in agent, it can fill in the quote form while you watch. From a chat app or a terminal, it can draft the enquiry with you and give you a link that opens our form already filled in. Either way, you check it and press send.

Agents can’t book a visit or change anything on the site. The tools only see what is already public here, and nothing about visitors is collected or shared through them. You can also use the contact page directly.

About WebMCP

“WebMCP” currently means two related things:

• The browser standard. A proposal from the W3C Web Machine Learning Community Group that lets a web page offer tools to the AI agent built into a browser. This is what option 1 uses. Proposal on GitHub
• The open-source bridge. A JavaScript library that connects a web page to desktop AI apps through a local connection. This is what option 2 uses. GitHub · npm

Both are built on the Model Context Protocol, the open standard AI apps use to talk to outside tools and data.