{"path":"/guide/how-it-works/architecture","content":{"title":"Architecture overview","summary":"How Everything serves public pages and exposes an agent-readable wiki.","breadcrumbs":"[All topics](/guide) / [About Everything](/topics/about-everything) / [How Everything works](/guide/how-it-works)","body":"Everything is a dynamic wiki built on **wf-store**, an asset and retrieval service. Its public interface interprets stored assets as pages, collections, and templates rather than maintaining a separate content database.\n\n## A page read\n\nA browser or MCP client asks Everything for a public page. The application resolves the page's canonical route using its tags, retrieves the selected immutable source, and returns either a structured result or rendered HTML. Only published and otherwise eligible pages are exposed.\n\nThe browser displays a lightweight reading shell, then loads the article and related links separately. MCP clients can bypass the human interface and request page content, page lists, search results and related articles directly.\n\n## Content and presentation\n\nContent is stored as immutable assets, with a readable text rendition for retrieval. **Tags carry the mutable interpretation:** logical page identity, route, publication state, collection membership and chosen template. Templates are declarative layouts rather than arbitrary executable code.\n\nThere is **one corpus for the wiki**, containing the different asset types. A separate page-head registry and static HTML build step are unnecessary. [The tag model](/guide/how-it-works/tag-model) describes how page identity and visibility are interpreted.\n\n## Search and performance\n\nEverything provides bounded public browse and search operations, using the retrieval capabilities of wf-store. It reuses immutable source bytes and validated templates in warm application caches. Page selection still reflects stored metadata and may be temporarily stale because the underlying storage is eventually consistent.\n\n## Security boundaries\n\nPublic reading is anonymous and read-only. Privileged content maintenance is separately authorised; no public tool exposes raw unrestricted storage operations or unpublished page assets. [Security and ownership](/guide/how-it-works/security) explains these boundaries.\n\nThe key principles are [deterministic interpretation](/guide/how-it-works/tag-model), [repair over prevention](/guide/how-it-works/repair-philosophy) and an interface that is equally understandable to humans and agents.","navigation":"[How Everything works](/guide/how-it-works) · [The tag model](/guide/how-it-works/tag-model)","related":[{"title":"The tag model","url":"/guide/how-it-works/tag-model"},{"title":"Caching and freshness","url":"/guide/how-it-works/caching"},{"title":"wf-store","url":"/guide/how-it-works/wf-store"}]}}