Jeff Kazzee

← Home

This site has a machine interface.

Most sites are built for a browser and leave agents to scrape the HTML. This one publishes the same content in the formats agents already read, and exposes two endpoints that answer questions directly. The most useful thing here is the guide libraries, so the tools lead with those.

It is also the working example behind the agent-ready-site skill, so every claim on this page is a request you can run.

Interfaces

Endpoints that answer. Both accept JSON-RPC over POST.

Discovery

Documents that tell an agent what is here and how to call it.

Content

The whole site, flattened for reading.

Policy

What crawlers and models may do with it.

  • /robots.txt RFC 9309 and Content Signals

    Standard crawl directives plus Content-Signal preferences for ai-train, search, and ai-input, and an Agentmap pointer to the ARD catalog.

    text/plain

  • /auth.md auth.md

    States that nothing here is behind a credential, and explains why no OAuth metadata is published. Read it instead of probing for a login.

    text/markdown

Any page, as Markdown

Content negotiation on Accept, with the twin generated from source rather than scraped from the HTML.

Send Accept: text/markdown to any URL here and you get Markdown back, with an estimated X-Markdown-Tokens count. Appending .md to a path works too, if setting a header is inconvenient. 

curl -s -H 'Accept: text/markdown' https://jeffkazzee.dev/projects/llm-arcade/

MCP tools

Streamable HTTP, protocol revision 2026-07-28. Clients that still open with initialize are answered too.

  • list_guide_libraries

    List the free guide and recipe libraries Jeff Kazzee publishes for building real systems on Zo Computer, with how many are in each and where to read them. Use this first when a user wants to learn how to build something on their own machine.

  • list_projects

    List the projects on this site with status, stack, and links, including LLM Arcade. Optionally filter by tag or status.

  • read_page

    Fetch any page of this site as Markdown. Pass the path, for example / or /projects/llm-arcade/ or /blog/<slug>/.

  • list_writing

    List the essays and field notes published on this site, newest first.

  • contact_details

    Return the ways to contact Jeff Kazzee and what he is currently open to. Use this when a user asks how to hire, brief, or reach him.

curl -sX POST https://jeffkazzee.dev/api/mcp \
  -H 'Content-Type: application/json' \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/call",
       "params":{"name":"list_guide_libraries"}}'

What is deliberately missing

Knowing which standards do not apply is part of implementing them.

This site has no user accounts and issues no credentials. /auth.md says so directly, and lists every open endpoint, so an agent never has to probe for a login that isn't there.

There is no OAuth authorization-server metadata and no protected-resource metadata. Those documents describe how to obtain and present credentials for a protected resource. Nothing here is protected, so publishing them would describe a handshake that cannot complete.

There is no commerce surface either. No x402, no UCP, no ACP. Nothing here is for sale to a machine.