· Steven Mays · AI & Development  · 3 min read

I Built an MCP Server for My Site

A public, read-only MCP server at api.mays.co/mcp lets Claude, ChatGPT, Cursor, or VS Code run my retirement simulator, search Buffett's letters, and read my posts.

A public, read-only MCP server at api.mays.co/mcp lets Claude, ChatGPT, Cursor, or VS Code run my retirement simulator, search Buffett's letters, and read my posts.

I built a public MCP server for this site at https://api.mays.co/mcp. Connect it to Claude, ChatGPT, Cursor, or VS Code, and your assistant can run my retirement simulator, search Warren Buffett’s shareholder letters, and read every post on this site. It’s read-only and needs no sign-in.

MCP (Model Context Protocol) is the standard way an AI assistant calls tools that live outside it.

The MCP page has one-click install buttons for Cursor and VS Code, steps for Claude and ChatGPT, and a live inspector that sends real requests from your browser. The same steps are at the end of this post.

Here’s what your assistant gets:

  • simulate_retirement: the retirement simulator’s engine, run over 100,000 market histories.
  • search_buffett_letters: semantic search over 25 Berkshire Hathaway letters, 1998 to 2023.
  • list_posts and get_post: every post on this site, as Markdown.
  • about_steven: my background, stack, and links.
  • stress_test_retirement and ask_warren: two prompts, which most clients show as slash commands.

Ask your assistant whether a 4% withdrawal lasts 30 years with 60% in stocks. It calls simulate_retirement and gets back a 92.71% success rate: 7,286 of the 100,000 histories run out of money, and the median failure comes in year 26. The simulator page shows the same results, because the server and the page run the same 17 KB Rust engine compiled to WebAssembly.

No tool on the server asks a model to write anything. The server computes and retrieves, and your assistant writes every answer.

It runs as one AWS Lambda behind the same api.mays.co gateway as the site’s other tools. One stateless function is enough.

Connect your assistant

Claude Code

Run this in a terminal, then start claude:

claude mcp add --transport http mays-co https://api.mays.co/mcp

Claude (web and desktop)

  1. In Claude on the web or the desktop app, open Customize › Connectors › Add custom connector.
  2. Paste https://api.mays.co/mcp.
  3. Choose No sign in.

The Free plan allows 1 custom connector. On Team and Enterprise plans, an Owner must add it.

ChatGPT

  1. Turn on Developer mode in ChatGPT’s settings.
  2. Add an app and paste https://api.mays.co/mcp.

Developer mode is a beta for paid plans, on the web only. If the menu moved, search settings for Developer mode.

Cursor

Add to Cursor. Cursor opens and asks you to confirm.

Or add it by hand to .cursor/mcp.json:

{
  "mcpServers": {
    "mays-co": { "url": "https://api.mays.co/mcp" }
  }
}

VS Code

Install in VS Code. VS Code opens and asks you to install the server.

Or add it by hand to .vscode/mcp.json:

{
  "servers": {
    "mays-co": { "type": "http", "url": "https://api.mays.co/mcp" }
  }
}

Resources

Back to Blog

Related Posts

View All Posts »