llms.txt and catalogs

Falcata publishes several machine-readable indexes of the same content, each shaped for a different kind of consumer. This page maps out which file to fetch depending on what you're trying to do.

/llms.txt — a link index

A short, curated list of links into the site: a title, a description, a "when to use" note, and per-section link lists (- [Title](/path): description). Treat it like a table of contents written for an agent rather than a human skimming a nav bar.

Use this when you want to decide where to look before committing to reading anything in full — it's small, cheap to fetch, and refreshed regularly.

/llms-full.txt — the full text, concatenated

Every published page's content, concatenated into one file with FILE: <path> separators between them. This is the "just give me everything" option: no navigation, no summarizing, the actual page bodies back to back.

Use this when you want the whole public corpus in one fetch — for example, seeding a local index or answering a question that might touch several pages you can't predict in advance. It's larger and refreshed less often than llms.txt, so prefer llms.txt first if you already know what you're looking for.

/.well-known/ai-catalog.json — structured entity catalog

An Agentic Resource Discovery (ARD) manifest: a machine-readable list of named, typed entries (an identifier, displayName, type, optional description, tags, capabilities, and example queries) describing discrete things this site exposes — not pages, but resources or capabilities an agent might want to reason about as first-class objects. Each entry can also carry a trustManifest tying it back to a verifiable identity.

Use this when you want structured metadata about what this site offers rather than prose to read — for building a catalog of capabilities across multiple sites, for instance.

/.well-known/api-catalog — RFC 9727 linkset

A RFC 9727(opens in a new tab) API catalog, served as application/linkset+json. It links each API surface to its machine-readable description (service-desc, typically an OpenAPI document) and its human documentation (service-doc).

Use this when you're specifically looking for API definitions — it's the standard, IANA-registered way to ask "what APIs does this host expose, and where's the spec for each one" without knowing Falcata-specific paths in advance.

/.well-known/agent-skills/index.json — packaged skills

An Agent Skills index (schema agentskills.io/discovery/0.2.0) listing named, typed skill packages — each with a name, description, url to the skill content, and a content-addressed digest so a client can detect when a skill has changed without re-fetching it speculatively.

Use this when your agent runtime understands the Agent Skills format and you want to load a packaged, reusable skill rather than free-form page content.

Every page as markdown

Every public page is also served as plain markdown at the same path with .md appended (the homepage is /index.md). HTML pages advertise this via a <link rel="alternate" type="text/markdown"> tag in the head, so you can discover the markdown twin from the HTML you already fetched instead of guessing the URL. See Every page as markdown for details.

Choosing between them

See also: For agents, Every page as markdown, REST API.