Readable by Everything
A site that only humans can read is half-finished.
A site that only humans can read is half-finished. That’s the premise behind the last forty-eight hours of work — making every property I run legible to the things that aren’t people.
OG images fill the link preview. JSON-LD gives a crawler structured data: name, role, employer history, same-as links, the whole graph. llms.txt so an agent can read the site in one pass without parsing HTML. These are table stakes for a web where half the readers don’t have screens.
AgentNDX crossed 137 servers this week. Each one gets the same treatment: structured metadata, verified status, machine-readable descriptions. The directory is an API surface for discovery. When an agent asks “what MCP servers handle payments,” the answer comes back as structured data.
The pattern is the same at every scale. A personal site, a directory, a company page — the question is always: can an agent understand what this is? Machine-readable information lets the site participate in that network.
I’m building nodes.