AI-readable hosting does not mean opening every system to every crawler. It means publishing the right public information in the right formats and being strict about actions that change customer accounts.
Three files and interfaces matter most: sitemap.xml, robots.txt, llms.txt, and MCP.
Sitemap.xml tells systems what exists
A real sitemap gives search engines and AI answer systems a clean list of canonical public pages. For a hosting company, that should include product pages, pricing pages, domain pages, legal pages, support-style guides, and API documentation.
Without a sitemap, agents may miss important pages or rely on stale links.
Robots.txt declares crawler policy
Robots.txt is where a business decides what public crawlers may fetch. OneNet Servers allows public commercial and support content to be indexed while blocking sensitive write endpoints and training-oriented crawlers where supported.
That policy balances visibility with control:
- Public pages can be discovered
- Checkout and account endpoints stay blocked
- User-triggered assistants can fetch public information
- Training crawlers can be blocked if that is the business preference
llms.txt gives agents a short operating guide
llms.txt is a plain text guide for AI systems. It should explain what the business does, what public pages matter, what APIs exist, and what agents are not allowed to do.
For OneNet Servers, llms.txt points agents to product pages, pricing, domain search, API docs, and the MCP endpoint. It also states that direct WHMCS write access is not available.
MCP turns product discovery into tools
MCP gives agents a structured way to call tools. For hosting, the useful first tools are read-only:
- List products and pricing
- Search domain availability
- Explain checkout requirements
Write tools can come later, but they should be approval-gated. A tool that can create an order or register a domain needs customer approval, fraud controls, and WHMCS reconciliation.
The real goal
The goal is not to make the website "for AI" instead of humans. The goal is to make OneNet Servers easier for both to understand. Humans get clearer pages. Agents get structured public data. Customer-impacting actions remain controlled.
