What is specification.website?

The Website Specification is an open reference site that collects the rules a website should follow across HTML structure, accessibility, security, SEO, performance, privacy, and readability by AI agents. The web is built on many overlapping standards from different organizations, and this project pulls them into one place with every requirement traced back to its original source. It is not tied to any particular technology or framework: it describes what the outcome should be, not which tool to use to get there. The content is organized into categories covering areas like HTML document basics, search visibility, WCAG-aligned accessibility rules, security headers and transport policies, well-known URI paths, Core Web Vitals performance, privacy signals, resilience under failure, and internationalization. Each page states whether a requirement is required, recommended, optional, or something to avoid, and links to the relevant standards from bodies like the WHATWG, W3C, IETF, and IANA alongside search engine guidance. A separate category covers agent readiness, which means making a site's content structured and accessible enough for AI agents and tools to read and use correctly. This includes things like llms.txt, a file format for exposing content summaries to AI systems. The project is built with Astro and deployed to Cloudflare Pages. Running it locally requires Node.js 22.12 or later. The build derives several outputs automatically from the spec source files: a checklist page, sitemap, RSS feed, per-page Markdown endpoints, a Pagefind search index, and an MCP server that exposes the spec content to AI agents that support the Model Context Protocol. Those derived files should never be edited by hand. Contributions follow three stated rules: cite sources, stay platform-agnostic, and be honest about whether something is settled or contested. Code is MIT licensed, content is CC BY 4.0.