I'm Serhii Kravchenko.
I build agent systems.

What I have built and measured

The proof on this site is checkable, not claimed. A live connector anyone can try: add it to your own Claude in a minute, nothing to sign into, and book a real call with me. A census of the Claude connector directory, run by my agents every week inside Anthropic's terms and published as an open dataset. And field notes where every number was measured on my own server: which of what a server sends actually reaches the AI, what a chat card really weighs, what the platform silently drops.

Beside yesmcp the same team builds two products: avoidcontent, where I am co-founder and R&D lead, and site-dna-manager, my own product at the R&D stage. They are not the subject of this site, but they are why the notes here come from live systems rather than from reading documentation.

What that team and I have done for other companies, and for my own products, is on a separate page: the cases, department by department, each with what changed for the people who hired me.

The path, briefly

I have been building agent systems full-time for two years. Before that, eight years running my own e-commerce businesses on Amazon and Shopify. In these two years: acting Head of AI at Keaworld, a stablecoin banking infrastructure company with payment providers and iGaming operators among its clients (October 2025 to July 2026), agent solutions shipped across very different industries, and onboarding executives and teams onto agent systems. The vertical never mattered much to me. What matters is a challenge where an agent system can be pushed to its real potential.

Where I can be useful

• Your business inside AI chats. Customers ask, choose and book inside Claude and ChatGPT; an MCP connector is how a business gets there.
• Sales and marketing working by hand. Agent systems I have shipped for lead generation, outbound, content and SEO, customer support, HR and IT.
• Getting the team past demos. I bring teams and executives into agent systems past the surface level, on their own work.
• Problems with no known answer. R&D is my favourite work: the answer has to be found, not looked up.

How I work

I live in Cyprus and spend my days inside agent systems, usually two or three projects in parallel. The routine of yesmcp, monitoring, checks and daily operations, is run by a team of AI agents I built. Whatever I build, the code and the specification belong to the company from day one, the documentation is written so that a stranger can pick the system up, and nothing that reaches a customer goes out without a person saying yes.

The standard this site is held to

Everything published here is something I ran myself, and every number was measured on a live system. The prose is written with AI assistance, which seems worth saying out loud on a site about AI tooling. Where something is still a guess, it says so. If something here turns out to be wrong, I would rather hear it and correct it in place with a dated note than quietly edit the page.