AI and agent access

This documentation is built to be read by software as well as by people. Nothing here requires scraping rendered HTML.

The outputs

Artefact	Path	Use
Compact LLM index	/llms.txt	An orientation file: scope, core model, and where to look
Full LLM bundle	/llms-full.txt	The complete normative content in one file
Rendered LLM page	/llms.html	The same content, readable in a browser
Page manifest	/manifest.json	Every page with its stable ID, area, status, sources, and content hash
Raw Markdown	/raw/<page-id>.md	The exact source of any page
Specification	/specification.md	The compatibility specification as a standalone file
Source ledger	/sources.md	Every source, revision, and why it is used
OpenAPI documents	/contracts/openapi/*.json	Validated 3.1 contracts
JSON Schemas	/contracts/schemas/*.json	The shared component library
Conformance vectors	/conformance/vectors/*.json	Executed deterministic cases
Changelog feed	/changelog.xml	A subscribable feed of documentation changes
Sitemap	/sitemap.xml	Every route

Published OpenAPI documents

Document	Operations	Source revision	Download
Universe ARC-20 index and feed API Verified ticker resolution, token details, holders, and confirmed activity for ARC-20.	5	670030e50071	JSON
Universe Atomicals NFT, Realm, and Subrealm API Read model for plain NFTs, Realms, and Subrealms projected from Atomicals ElectrumX.	36	1d9109c68d95	JSON
Atomicals Marketplace v1 authority Four isolated protocol authorities for ARC-20, plain NFTs, Realms, and Subrealms.	35	670030e50071	JSON
Shared schema library One component library every document draws from	1	1.0.0	JSON
Source manifest Every consumed source and its revision	1	1.0.0	JSON
Conformance vectors Executed ARC-20 allocation cases	1	1.0.0	JSON
CLI inventory Commands parsed from the pinned CLI source	1	1.0.0	JSON
AIP registry Proposals with implementation evidence, never assumed	1	1.0.0	JSON
AVM opcode inventory Generated from the beta interpreter source	1	1.0.0	JSON
Ecosystem registry Records with evidence and last-verified dates	1	1.0.0	JSON
Page manifest Every page with its stable ID, status, sources, and content hash	1	1.0.0	JSON

What the LLM files guarantee

1. Normative sources come first. The pinned validator revision and the executed vectors lead; explanatory prose follows.
2. Proposals and experiments are labelled. Nothing that is proposed or beta is presented as live behavior.
3. Protocol and product are never collapsed. Every claim says whether it describes the protocol or a Universe service.
4. Every claim is dated and pinned. The source revision and verification date travel with the content.

Page identity

Every page has a stable pageId recorded in its frontmatter and exposed in the manifest, in a bu:page-id meta tag, and in the raw Markdown path. Page IDs are never reused and never renamed without a redirect.

Use the page ID as the key, not the URL. URLs can move between locales and versions; a page ID cannot.

Content hashes

The manifest records a hash of each page’s content. An agent can detect a change without diffing text, and a translation records the hash of the English source it was made from, which is how the site marks a translation as stale.

Fetching politely

robots.txt allows crawling. Two requests:

1. Read manifest.json once and fetch only the pages whose hash changed.
2. Prefer /raw/<page-id>.md over the rendered page. It is smaller and it is the source.

• Documentation MCP server

  Query pages, contracts, and vectors locally, with no network access and no credentials.

• Contracts

  OpenAPI documents, JSON Schemas, and the source manifest.