I build AI agents and the backends they run on.
Lead AI Engineer. 15+ years in production systems.
MCP server
Send request calls api.mays.co from your browser
One public endpoint gives your AI assistant the retirement simulator, Warren Buffett’s letters, and every post on this site. Read-only, no sign-in.
https://api.mays.co/mcp tools/call simulate_retirement
Computed when this site was built
calling…
%
No simulation came back from api.mays.co. Try it in the inspector.
Request from your browser
> POST https://api.mays.co/mcp
{ "jsonrpc": "2.0", "id": 1, "method": "tools/call", "params": { "name": "simulate_retirement", "arguments": { "withdrawal_rate_percent": 4, "years": 30, "stock_share_percent": 60 }, "_meta": { … } }}
Response from api.mays.co
< Not sent
"structuredContent": { "success_rate_percent": , "paths": , "failed_paths": , "median_failure_year": , …}
Ran out of money, by year
< Projects
How I build
I’ve shipped 10+ production agents as state machines with per-state retries and cost ceilings.
Related: Slow Agents and Spare Tokens: Building an Investing Agent That Optimizes for Being Right
My event-driven pipelines absorb failures at thousands of events per second, and my serverless migrations cut infrastructure cost by 50%.
Related: Setting Up An API Subdomain - Solving CORS Issues and Production AI Reliability
I’ve built 3 MCP servers. One is public on this site for your assistant to use.
Try it on the MCP server page