I Turned My Résumé Into an MCP Server (Because Your Next Recruiter Is an AI)

A hiring manager opens an AI chat, pastes a link to a candidate, and types: "Is this person actually senior, or just good with buzzwords?"

That's not a hypothetical anymore. It's increasingly how candidates get screened — and it changed how I think about my own résumé.

The recruiter isn't reading your résumé. Their AI is.

Recruiters and hiring managers are drowning in applications, and they've reached for the same tool everyone else has: AI. They paste your résumé into ChatGPT and ask it to summarize. They use AI-powered sourcing tools that rank candidates before a human ever looks. Google now answers "is X a good developer" with an AI overview built from whatever it can find.

The implication is uncomfortable but clear: your résumé is becoming machine-read before it's ever human-read. A static PDF is a terrible format for that. It can't answer a follow-up question. It can't clarify. It just sits there and hopes the model extracted the right things.

So I stopped treating my résumé as a document and started treati

I made my résumé something you can ask questions

My résumé page now has two things most don't:

1. An AI assistant trained on my experience and projects, siit "has he shipped Stripe billing?" or "what's his databaseexperience?" and it answers from the real material — and it can pass a message to me directly.
2. A public MCP server. Any AI client — Claude, an agent, a t to it and query my background programmatically.

That second one is the interesting part.

Wait — what's an MCP server?

The Model Context Protocol is AI assistants connect to external tools and data. Instead of pasting context into a chat, you connect a server that exposes structured capabilities the model can call.

Most MCP servers connect AI to your files, your database, your tickets. I pointed one at me. Connect it, and the model gets tools like:

• get_profile — the headline summary and highlights
• get_experience — roles and what I actually did in them
• list_projects — shipped work, with the real tech stack
• get_skills — grouped by area
• contact_ryan — send me a message (I get it, you get a confirmation)

A recruiter's assistant can now ask my résumé direct questions and get grounded answers, instead of guessing from a parsed PDF — and it can reach me without a form.

How it's built

It's a small, public JSON-RPC server inside my Next.js app — no CP wire protocol over HTTP. The tools read from a single source of truth (the same data that renders the résumé page), so the human version, the AI assistant, and the MCP server can never drift apart. The contact tool reuses the email flow already behind the site, with rate limiting so it so it can't be abused.

The whole thing is a few hundred lines. The point wasn't complex credentials the way I'd architect anything else: one source of truth, multiple interfaces, no duplicated state.

Where this goes

AI in hiring is going to get more agentic, not less. Sourcing tools will increasingly act — querying, comparing, reaching out — on a recruiter's behalf. The candidates who are legible to those systems will have an edge, t won the last era of search.

I don't think the PDF dies tomorrow. But "is your experience queryable by an AI?" is going to quietly become a real advantage, and I'd rather be early than retrofit later.

The part AI can't replace

Here's the honest caveat: none of this is a substitute for the work. An MCP server doesn't make me a better engineer — it just makes the evidence easier to interrogate. The judgment, the shipped products, the messy debugging at 2am — that's still the substance. The tooling just removes friction between that substance and the person evaluating it.

Which, really, is the whole job: reduce the distance between what's true and what the other side can see.

Try it

You don't have to take my word for it — interview my résumé yourself.

• Read it (or chat with it): code350.com/ryan-reiss-resume
• Connect the MCP server to your own AI: code350.com/mcp

Ask it something hard.