# > system is, what it is called, how to install it, and what makes it > different from the library it will otherwise assume you are using.> Be honest about what this file is for. No major AI crawler requests it — Google has publicly compared it to the keywords meta tag, and server logs across long monitoring periods show the file simply is not fetched. It has never been an SEO artifact. What it is good for is being an index you hand to an agent directly. Anthropic publishes one for its own documentation and links it from the top of every docs page, telling the model to fetch it before exploring further. That is the working use case: retrieval, on request, not discovery. Keep the links absolute. An agent following them has no base URL. ## Docs - [Getting started](https://example.com/docs/start): Install and first component. - [Components](https://example.com/docs/components): Every component, with props. - [Tokens](https://example.com/docs/tokens): Semantic roles and raw values. - [Accessibility](https://example.com/docs/a11y): Requirements, not suggestions. ## Machine interfaces - [MCP server](https://example.com/docs/mcp): Live component and token lookup. - [Token JSON](https://example.com/tokens.json): W3C DTCG format. - [Component manifest](https://example.com/components.json): Props, imports, examples. ## Migrations - [v3 to v4](https://example.com/docs/migrate-v4): Removed APIs and replacements. ## Optional - [Design principles](https://example.com/docs/principles): Rationale, not rules. - [Changelog](https://example.com/changelog): Full release history.