specification.website
Joost de Valk's platform-agnostic website specification maps the web foundations, well-known URIs, and agent-readiness signals Open for Agents helps WordPress sites expose.
References
Open for Agents sits inside a wider web ecosystem. These references describe the standards, drafts, validators, and technical thinking that shape agent-readable WordPress publishing.
Joost de Valk's platform-agnostic website specification maps the web foundations, well-known URIs, and agent-readiness signals Open for Agents helps WordPress sites expose.
The agent-readiness section defines the practical site signals agents rely on, including stable URLs, structured data, robots controls, and machine-readable endpoints.
The MCP specification defines the protocol requirements for clients, servers, tools, resources, transports, and authorization behavior.
The webmachinelearning/webmcp repository hosts the WebMCP draft for exposing JavaScript and form-based tools to browser agents while preserving page context and user control.
RFC 8615 defines the /.well-known/ URI pattern used by many discovery, identity, authorization, and agent-readiness endpoints.
The IANA registry is the authoritative index of assigned /.well-known/ names and their defining specifications.
RFC 9727 defines /.well-known/api-catalog for automated discovery of published APIs and their machine-readable metadata.
RFC 9264 defines the Linkset format used by API catalogs to publish typed links in a machine-readable document.
RFC 8414 defines /.well-known/oauth-authorization-server, a discovery endpoint used by OAuth-based MCP authorization flows.
RFC 9728 defines /.well-known/oauth-protected-resource, which lets clients discover how an OAuth-protected resource expects to be accessed.
This active Internet-Draft proposes an mcp: URI scheme and a /.well-known/mcp-server discovery path for finding MCP servers.
This active Internet-Draft defines a directory for HTTP message-signature keys, relevant to Web Bot Auth and authenticated automated traffic.
Cloudflare's scanner checks public URLs for agent-readiness signals across discoverability, content access, bot controls, protocol discovery, and commerce.
Cloudflare URL Scanner reports include Agent Readiness scoring for public URL scans, with detail across emerging AI and automated-discovery signals.
Joost de Valk's retrofit case study shows how markdown alternates, API catalogs, agent skills, MCP, WebMCP, and OAuth tradeoffs show up on an existing site.
Joost de Valk's case study shows how source content, markdown endpoints, MCP, and discovery headers fit together on a site designed for agents from the start.
This post argues for implementing machine-readable web standards before broad crawler adoption makes their value obvious.
This background post explains why specification.website exists and why agent-readiness belongs with ordinary website fundamentals.
Anthropic's Claude API documentation explains direct connections to remote MCP servers, tool configuration, OAuth tokens, and current connector limits.
Chrome's WebMCP guide explains the proposed browser APIs, local testing flow, limitations, and testing tools for client-side agent tools.