Making a Developer Portfolio Agent-Ready

Recruiters still browse HTML. Agents increasingly read link headers, llm.txt files, OAuth discovery metadata, and browser-registered tools. A portfolio that speaks only to humans leaves signal on the table.

I treated erollmaxhuni.com as a small research project: implement the specs, run scanners, and document what actually passes versus what is theater.

What agents look for first

Plain-language profiles (llm.txt), API catalogs (RFC 9727), Auth.md registration metadata, and optional WebMCP tools on page load. Each layer answers a different question: who is this person, what endpoints exist, how do I register, what can I do interactively?

The interesting finding: consistency beats volume. Contradictory metadata — free access in one file, payment required in another — fails validation faster than missing features.

What humans still need

Agents do not hire you. Hiring managers do. Agent-ready markup is additive SEO for machines, not a substitute for case studies, contact paths, and proof of production delivery.

The sustainable pattern: one data source (case studies, stack, contact) rendered for HTML, markdown negotiation, and discovery JSON alike.