Vlad's Personal Website

Inspiration

I wanted my personal website to work like a chatbot. Instead of scrolling through a portfolio, people could ask questions and learn about the parts of my work that actually interest them.

WebMCP takes that idea one step further: you don't even need to ask the questions yourself. Your agent already knows something about you from its memory. Why not let it interview my agent and find out what we have in common?

What it does

haltakov.com lets people learn about me through a conversation. With WebMCP, a visitor's agent can have that conversation on their behalf.

The agent can use what it knows about its user to ask relevant questions, follow up on interesting answers, and produce a personalized report highlighting shared interests and reasons to connect.

Someone building an AI product might discover relevant projects. Someone interested in computer vision might explore my engineering background. The same website can lead to very different conversations.

How we built it

The foundation is Chatbot Page, my open-source project for building conversational homepages. It provides a reusable React package and a Next.js example, with support for curated Markdown answers, streamed LLM responses, and retrieval from your own documents.

The WebMCP integration adds two tools: ask_vlad for a multi-turn interview and show_interest_summary for presenting a personalized report. The visiting agent decides what to ask based on its user's interests, rather than following a fixed questionnaire.

Challenges we ran into

A key design challenge is making the conversation useful to two different audiences: people chatting directly and agents asking questions for someone else. An agent needs room to follow up and explore, but the result still needs to be useful to the human behind it.

Another important boundary is personal context. The visiting agent can use its memory to choose questions without handing that entire memory to my website. The site receives the questions it submits, so those questions still matter from a privacy perspective.

Accomplishments that we're proud of

The project brings agent-to-agent conversations to an actual personal website. It turns a generic introduction into an interview tailored to the visitor.

The chatbot foundation is also open source, so other people can build conversational homepages for themselves or their companies.

What we learned

The interesting part isn't just letting an agent operate a website. It's letting that agent bring context to the interaction.

My agent knows about me. Your agent knows about you. A conversation between them can uncover relevant connections that a static biography leaves the visitor to figure out.

What's next for Vlad's Personal Website

A natural next step is to explore this beyond personal introductions: discovering potential collaborators, finding relevant expertise, or preparing for a meeting.

The idea stays simple: let your agent talk to mine, then tell you what's worth knowing.