railway serves an AI Catalog
manifest at /.well-known/ard.json on railway.com, declaring
8 agentic resources
that a discovery service can index without asking anyone’s permission.
Run against this exact body with the project’s own
conformance CLI.
Discovery path
ard.json
Served at /.well-known/ard.json — the path ARD v0.91 says a consumer MUST fetch.
Declared spec version
0.91
The specVersion in the manifest. 1.0 is current for the AI Catalog data model; anything lower is an earlier draft. ARD reads this field as transport-defined and ignores it.
⚠ Hard checks failed. An ARD client cannot assume this manifest parses to the model the specification defines.
no-host-object
Optional affordances missing. None of these break a client, but each one costs the publisher search relevance or verifiability.
content-type-not-ai-catalog-json
cors-not-readable-cross-origin
entry-no-representativeQueries
8 entries
6 media types
17 representative queries
8 trust manifests
Host identifier: none
What this manifest advertises
Railway Remote MCP Server
urn:air:railway.com:server:railway-mcp
Deploy and operate apps, databases, volumes, and cron jobs on Railway from any MCP client.
“show me the logs for my failing Railway deployment”
“set an environment variable on my Railway service”
Railway Public API (GraphQL)
urn:air:railway.com:api:public-graphql
Railway's public API — a single GraphQL endpoint with the schema available by introspection. Authentication and token types are described at https://railway.com/auth.md.
OpenAPI 3.1 description of Railway's HTTP surface: the GraphQL entrypoint with its bearer/OAuth2 security and named scopes, plus the OAuth 2.0 / OIDC endpoints.
How the Railway API behaves under failure: the error convention and code catalog, rate limits and headers, retry/idempotency guidance, and the versioning/deprecation policy.
The four ways an agent or automated client holds a Railway credential — CLI login, remote MCP over OAuth, API tokens, and Login with Railway — with discovery, claim, error, and revocation walkthroughs.
The publishing guide asks for three things at the transport layer, because a crawler is the
consumer that matters and a crawler is a browser-shaped client.
HTTPS only — yes (HTTP 200)
Content-Type: application/json —
yes
Access-Control-Allow-Origin: * —
no — header absent
Evidence. Fetched https://railway.com/.well-known/ard.json on 2026-09-13, HTTP
200. The verbatim body is stored alongside its manifest in the
railway repository
as ai-catalog/railway-ai-catalog.json. Nothing on this page is derived
or generated: an AI Catalog manifest is served from the publisher’s own domain or it does not
exist, which makes it — like an
A2A Agent Card — one of the few agent artifacts that cannot be
produced on a provider’s behalf. Grades are recomputed on every build; the specification is a
v0.9 draft and
these verdicts will move when it does.
Every descriptor here is available over the APIs.io API and to AI agents over MCP. Agent Discovery is not yet its own endpoint on the v1 API. Reach it through catalog search and the tag graph, or the MCP server.
MCP server
One button, every client — Claude, Cursor, VS Code and the rest.
https://apis.io/mcp
Tools for agent discovery
3 MCP tools reach this
apis_io_searchSTART HERE — APIs, providers and tags for one query, each with its total.
resolveTurn a domain, URL or GitHub org into the provider it belongs to.
find_cohortsEvery scored population of providers in the catalog.
Calls without X-API-Key need no key. Calls with it need a paid plan — get a key.
Get an API key
Free tier, no form to fill in. Signing in shares your email address with us — we
store it to create your key and to recognise you if you sign in with another
provider. See our Privacy Policy and
Terms.