Serve at least 500 characters of meaningful homepage content in raw HTML. Add a clear H1, keep deeper heading levels sequential, and remove excessive non-content markup.

Declare scoped API permissions where machines can read them: named OAuth scopes in your OpenAPI security schemes, or scopes_supported in RFC 9728 protected-resource metadata. Prose descriptions of roles help humans, but agents need the machine-readable declaration to request least-privilege access.

Keep the correct HTTP 404 status. The remaining requirement is a Markdown error body when agents request Accept: text/markdown. Include at least 20 characters explaining the error and a link to your docs, sitemap, or llms.txt. Verify with curl -sS -L -i -H 'Accept: text/markdown' https://yourdomain.com/some-path-that-does-not-exist. Check both the final 404 status and the Markdown body with Content-Type: text/markdown. Checking the status alone does not verify the missing requirement.

Tell agents when to reach for you: add a 'when to use this' section to your llms.txt (or a dedicated agent-instructions file) that names your best-fit use cases and how an agent should call you. Be specific about the jobs you are right for - generic marketing copy does not read as guidance.

Offer a free tier or trial, self-serve API key generation, and a sandbox environment. Agents can't fill out 'contact sales' forms.

Keep authentication enabled. Verify tool listing with an authorized MCP client. This unauthenticated scan cannot establish whether protected tools are missing or unusable.

Return standard rate-limit headers on your API responses (the RFC RateLimit headers, plus Retry-After on a 429) so agents can self-throttle in real time, and document the conventions alongside your API.

Document your error responses in your OpenAPI spec: give 4xx and 5xx responses a typed error schema (or use RFC 9457 application/problem+json). A consistent error object with a machine-readable code and a human-readable message lets agents handle failures without guessing.

Define typed JSON response schemas for every endpoint in your OpenAPI spec. Agents rely on these to know what fields they will get back; missing or partial schemas force trial-and-error.

Make your API spec self-describing: a unique operationId and a description on every operation, typed parameters, and response schemas. For GraphQL, a fully typed schema with a documented cost or rate limit reads best.

Give your MCP server-card (at /.well-known/mcp/server-card.json) a display name, an icon or logo, and a description - all three together are what reads as a complete, branded listing agents can present.

Let agents fetch markdown by appending .md to page URLs. Required for any credit: serve a markdown homepage at /index.md. For full credit (2/2): also serve a .md twin for each content page (e.g. /docs/auth -> /docs/auth.md). Content-Type should be text/markdown and the body should start with a top-level heading (not HTML).

Format your llms.txt as a navigation index: start with a markdown heading, include markdown links to deeper resources, and keep it under 30,000 characters. If you have more to say, move long-form content into /llms-full.txt or per-section files (e.g. /docs/llms.txt, /api/llms.txt) and link to them from the main index.

Publish an API catalog at /.well-known/api-catalog per RFC 9727. Serve it with Content-Type: application/linkset+json;profile="https://www.rfc-editor.org/info/rfc9727" and include a 'linkset' array with item entries pointing to your OpenAPI specs and service descriptions.

Publish RFC 9728 protected-resource metadata at /.well-known/oauth-protected-resource on your resource server (the host that actually serves the API, e.g. api.<apex>) with resource and authorization_servers. Publish RFC 8414 authorization-server metadata at /.well-known/oauth-authorization-server on the AS origin, and include the WorkOS auth.md agent_auth block with identity_endpoint, identity_types_supported drawn from the spec enum (anonymous, identity_assertion, service_auth - the assertion variant, the ID-JAG URN urn:ietf:params:oauth:token-type:id-jag, belongs inside identity_assertion.assertion_types_supported, not at the top level), and the identity_assertion.assertion_types_supported block when you advertise that type, so agents can check their assertion shape is accepted before minting. Cross-link by listing the AS origin in PRM authorization_servers, and point agent_auth.skill back at your published /auth.md. Spec: https://github.com/workos/auth.md.

Expose in-page tools through WebMCP, a proposed web standard for browser agents. Register tools with document.modelContext.registerTool() and use navigator.modelContext only as a trailing compatibility fallback. Declarative forms with toolname and tooldescription provide server-rendered evidence, but remain a preview and should not be your only tool surface. Chrome's origin trial covers versions 149-156, with shipping currently targeted for 157. ChatGPT can discover and call WebMCP site tools in the desktop app's built-in browser when the feature is available.

For long-running operations, return 202 Accepted and point agents at where to poll for the result (a status/location reference plus a job identifier in the body), documented in your OpenAPI spec, so work that does not finish in one request is still followable.

Publish official SDK packages across multiple language ecosystems (npm, PyPI, Go modules, RubyGems). Auto-generate them from your OpenAPI spec using tools like openapi-generator. For each package set the project URL or homepage to your product domain (package.json repository/homepage, PyPI Home-Page or project_urls, RubyGems homepage_uri) - this is how agents verify the package is your official SDK.

Publish a server card for each MCP server (draft MCP server-card extension). List each card in /.well-known/ai-catalog.json, or serve it at <server URL>/server-card. /.well-known/mcp/server-card.json is still read for a site with one server, but the extension does not recommend it. Include name, description, and version, and put the server's exact address in remotes[].url (serverUrl is also read) so agents match each card to the right server before connecting. Follow the extension schema for full credit: $schema set to https://static.modelcontextprotocol.io/schemas/v1/server-card.schema.json, a reverse-DNS name such as com.example/mcp, and no tools[] (agents always trust the live tools/list). Keep the description to 100 characters or fewer, as the schema requires. A card without that $schema is still read, and then an optional tools[] list completes it.