Skip to content
vladislav.dev

Service

AI integration and MCP server development

I add AI to products and internal tools: features built on large language models, agents that take actions, and MCP servers that let assistants such as ChatGPT and Claude work with your own data and systems, with the same permissions your people have.

Ask about this

Working prototype in about a week, then a fixed price

Who it's for

Product teams and agencies who want AI on their own data, not a generic chatbot, and need it to be secure, testable and maintainable.

What you get

  • A short scoping call and a written plan: what the AI should do, with which data, and how we will know it works
  • LLM features such as summaries, classification, search and chat over your data
  • MCP servers that expose your tools and data to AI assistants, behind sign-in and permissions
  • Checks on the AI output against known answers, so quality is measured, not guessed
  • Deployment on your infrastructure, with the code and the accounts in your name

How it works

  1. 01

    Scope and price in writing

    Before I start, you get the scope, a fixed price and a date.

  2. 02

    A demo every few days

    You see working software early and often, on a live preview link.

  3. 03

    Fast, and checked

    I build with AI coding agents, so I move quickly. Every change goes through tests and a review before it ships.

  4. 04

    You own everything

    The code, the hosting and the accounts are yours at the end.

Questions

What is MCP?

The Model Context Protocol is an open standard for connecting AI assistants to tools and data. An MCP server lets Claude, ChatGPT and other assistants read from and act on your systems in a controlled way.

Which AI models do you work with?

Mainly OpenAI and Anthropic Claude models, chosen per task on quality and cost. The code keeps the model swappable.

Is our data safe?

Your data stays in your systems. Access goes through sign-in and the same permissions your team has, and both OpenAI and Anthropic do not train on data sent through their APIs by default.

Start a project

Tell me about your project in two sentences.

I'll reply with how I'd approach it, a price and a timeline, usually within one working day.

vladislav.chiorescu@gmail.comLinkedInGitHub