Agent discovery

An agent that lands on your domain looks in a few well-known places before it reads any page. Nimbus publishes all of them by default. For everything an agent can read once it has found the site, see Docs for agents.

Path	What it is	Ships when
`/.well-known/ard.json`	Discovery manifest	Always, unless discovery is off
`/.well-known/ai-catalog.json`	The same manifest, under a second name	Always, unless discovery is off
`/.well-known/api-catalog`	API catalog (RFC 9727)	The site has an OpenAPI reference, unless discovery is off
`/.well-known/agent-skills/index.json`	Agent Skills index	The project has a `skills/` folder
Homepage `Link` header	Pointers to the files above	Always, unless discovery is off
`/robots.txt`	Crawler rules and Content Signals	`site` is set
`/llms.txt`	Index of the docs’ Markdown versions	Always

Discovery manifest

Paths: /.well-known/ard.json and /.well-known/ai-catalog.json · Type: application/json · CORS: Access-Control-Allow-Origin: *

Lists /llms.txt, the homepage Markdown at /index.md, each visible published API spec, the Agent Skills index, and the remote MCP server’s card when the site runs one. The two files have the same content. If your project already has a file or route at either path, the build fails until you move it, unless discovery is off.

API catalog

Path: /.well-known/api-catalog · Type: application/linkset+json · Standard: RFC 9727

One entry per visible API reference and version. Each entry links the reference page, its Markdown and its published spec, for example /api/openapi.json. The file has no extension, so a host that ignores _headers needs its own rule for the media type. See Published spec to keep a spec unpublished.

Agent Skills index

Path: /.well-known/agent-skills/index.json · Type: application/json

Lists each skill in the project’s skills/ folder, with a SHA-256 digest per file. See Publish Agent Skills.

Homepage `Link` header

Applies to: request-rendered homepages, and static builds on hosts that read _headers, including Cloudflare

Advertises the discovery manifest (rel="ard"), the API catalog (rel="api-catalog"), and each API family’s default reference and spec. Your own _headers rules win: a header they already set for a path is left out of the generated rule. On hosts without _headers, set the same headers in the host’s own configuration.

Content Signals

File: src/pages/robots.txt.ts · Default: Content-Signal: ai-train=yes, search=yes, ai-input=yes

Tells crawlers how they may use your content: training, search indexing, and answering questions. Edit the values in the starter’s route to match your policy, and if you set one to no, add the policy text from contentsignals.org above it. Signals are requests, not access control; private docs need authentication.

`agents.discovery`

Type: boolean · Default: true

Set to false when another site owns the domain, such as docs at example.com/docs beside a main site at example.com:

astro.config.tsts

const nimbusConfig = defineNimbusConfig({
  // …
  agents: { discovery: false },
});

Removes the discovery manifest, the API catalog, their Link header entries, and their _headers rules.
Keeps llms.txt, Markdown versions, Markdown negotiation, the remote MCP server, and Agent Skills.
Keeps your own files in public/.well-known/.

After switching, purge the cached homepage, since its Link header still lists the old files.

Under a base path

Discovery files belong at the origin root. With a base such as /docs, the Cloudflare adapter puts them there for you. On other hosts, deploy the emitted .well-known folder at the root beside the site:

host-root/
├── .well-known/   # discovery and Skills files
├── docs/          # the site
└── _headers       # when the host supports it