I turned my portfolio into an MCP server — so AI agents can hire me

My portfolio gets decent traffic. But lately, the first contact with a potential client increasingly isn't a person scrolling my site — it's their AI assistant, tasked with “find me an AI engineer under $2,000.” A normal portfolio is invisible to that flow. So I made mine visible on purpose.

The idea in one sentence

Expose my real professional profile — shipped products, fixed pricing, availability — as tools on a Model Context Protocol server, plus one gated write tool that lets an agent file a project brief on the founder's behalf. The result lives at mcp.djaouad.tech, and the code is open source.

The five tools

Four read tools (get_profile, search_projects, get_pricing, get_next_slot) give any MCP client — Claude, Cursor, ChatGPT connectors — grounded answers about who I am and what I charge, with demo links instead of marketing claims. One write tool, submit_project_brief, converts the conversation into action: a structured brief lands in my intake pipeline and my inbox.

The security pattern analysts actually recommend

Enterprise research on MCP keeps repeating the same advice: start read-only, put every write behind explicit gates, log everything. That's exactly how HireMe is built — reads are open, the single write is rate-limited per IP, briefs are reviewed by a human before anything happens, and no secrets sit anywhere near the endpoint. When a CTO asks “is this safe to connect?” the honest answer is yes, and the architecture backs it up.

Distribution compounds

After building it, I published the server manifest to the official MCP Registry under io.github.djoudad292/hireme-mcp, then to Smithery and Glama. Those listings feed each other — crawlers picked the listing up within a day. Every developer browsing those registries, and every agent platform reading them, now discovers a working demonstration of exactly the skill they'd hire me for.

What it changed

Three things. First, cold outreach gained a killer line: “don't take my word for it — point your AI at this URL.” Second, the server became my strongest case study; anyone technical can inspect the transport, tool schemas and rate limiting as proof of craft. Third, and strangest: my pipeline started receiving briefs written by models, summarizing what their humans need — often clearer than what the humans would have typed.

If you sell services, think agent-native

Websites made businesses legible to search engines. Profiles made people legible to networks. The next legibility layer is being legible to agents: structured, verifiable, actionable data your clients' AI can read and act on. You don't need my exact gimmick — you need your offer exposed somewhere machines can evaluate and act on it. The earlier you are, the less competition that layer has.

Want yours? I build agent-presence systems for service businesses — see the industry pages or grab a slot.