YULLAB — AI AGENT & MCP RESEARCH LAB

I research how
AI agents connect
with each other

Yullab documents the design and implementation work behind becoming an MCP expert. AI Agents are explored and learned together with other researchers.

Get in touch → Read the notes

MCP connection structure · example

The MCP client runs inside a host application (e.g. Claude Desktop) and talks to the server over a standard protocol.

§01 · Research

What I research

MCP Protocol Research

Centered on the Model Context Protocol — researching standardized ways to connect models with tools and data sources.

AI Integration Consulting

Designing and validating architectures that safely attach agents and MCP servers to existing systems.

AI Agent Development

Designing and building agent architectures that reason, call tools, and get goals done on their own.

§02 · Approach

How I build

01

Explore

Define the problem, the constraints, and the boundaries of the existing system.
02

Prototype

Quickly build a working agent/MCP integration into something testable.
03

Validate

Check stability, reliability, and failure handling against real workflows.
04

Integrate

Connect it safely into how the existing system runs, and document it.

Each step builds on the result of the one before it.

Example

How an MCP tool call works

When a client calls a tool registered on the server, the server returns the result in a standardized format. The example below is illustrative.

tools/call — example

// client → server
"method": "tools/call",
"params": {
  "name": "search_shipments",
  "arguments": { "port": "KRPUS" }
}

// server → client
"result": {
  "content": [{ "type": "text", "text": "Found 3 results" }]
}

§03 · Ecosystem

Clients that support MCP

Yullab's research is validated against the real MCP client ecosystem below.

Claude Desktop

Claude Code

Cursor

VS Code

Windsurf

Cline

§04 · Collaborate

Recording and sharing

Yullab is where I record and share my path toward becoming an MCP expert. I don't start with answers — only what I've designed, built, and verified myself stays here. I don't research AI Agents alone, though. If you're a developer or researcher who wants to try things out and learn together, feel free to reach out.

§05 · Contact

Let's talk about AI integration

Whether you're evaluating AI integration or have a topic you'd like to research together, feel free to reach out.

de.akirus@gmail.com

I research how AI agents connect with each other