Loopwise Docs
Concepts

Agent discovery

How an AI agent lands on a Loopwise storefront and discovers everything it can read and do — MCP, OpenAPI, Agent Skills, and markdown content — from predictable per-origin URLs.

Every Loopwise storefront advertises its agent-facing capabilities at predictable, per-origin URLs, so an AI agent can discover what a school offers in a single fetch instead of guessing. All URLs below are served per storefront origin — substitute the storefront's host (a custom domain like https://funi.best or a default subdomain).

Start here: the AI Catalog

The single entry point is the Agentic Resource Discovery catalog:

GET https://school.example/.well-known/ai-catalog.json

It returns one manifest listing every agentic resource on the origin — the MCP server, each Agent Skill, the OpenAPI spec, the ACP merchant endpoint, and the API catalog — each with a domain-anchored urn:air identifier, a media type, and a URL. An agent reads this first, then follows the entries it needs.

What the catalog lists

ai-catalog.json contains one entry per agentic resource on the origin:

Catalog entryURLWhat it is
MCP server/.well-known/mcp/server-card.json (transport at /mcp)Anonymous, read-only Model Context Protocol tools over the catalog.
OpenAPI spec/openapi.jsonOpenAPI 3.0 for the content and OAuth/OIDC endpoints. See Storefront content API.
Agent Skills/.well-known/agent-skills/index.jsonPublished skills for navigating and evaluating the catalog.
API Catalog/.well-known/api-catalogRFC 9727 linkset of the public APIs on this origin.
ACP merchant/.well-known/acp.jsonAgentic Commerce Protocol merchant discovery.

The catalog also carries a host.documentationUrl pointing at the storefront's llms.txt.

These aren't catalog entries, but they live at predictable URLs on the same origin and are worth following:

ResourceURLWhat it is
llms.txt/llms.txt, /llms-full.txtHuman/agent-readable index of the storefront's content and developer resources.
OAuth / OIDC/.well-known/openid-configurationPer-tenant identity provider — see OAuth quickstart.

Why per-origin

Each storefront is a distinct tenant. Identifiers and URLs in every document above resolve to the requested host, so a custom domain and a default subdomain each return a self-consistent view. The shape of these resources is identical across all schools (it comes from one platform); only the host and the underlying data differ.

Reachability

The machine endpoints are always served from the storefront origin, so an agent can operate even if this documentation site is unavailable. This page explains how the surfaces work; the per-origin documents are the source of truth for a specific school.

On this page