UZCBU MCP Server

An Model Context Protocol (MCP) server that connects the official exchange rates of the Central Bank of Uzbekistan (CBU.uz) to any MCP-compatible AI assistant (Claude Desktop, Claude Code, and others).

MCP is an open standard that emerged in 2024-2025 and by 2026 is becoming widespread in the AI world. It is used to connect AI models to external data and systems. This project is one such server for Uzbekistan.

What it can do

• get_all_rates — returns the official rates for all currencies for today

• get_exchange_rate — returns today's rate for a single currency (e.g., USD)

• get_historical_rate — returns the rate for a specific past date

• convert_currency — performs a conversion between two currencies based on the official rate (e.g., calculates how many EUR 100 USD equals)

Data source: the official open JSON API of cbu.uz. No API key or registration is required.

Related MCP server: Currency MCP Agent

Installation

git clone https://github.com/<sizning-username>/uzcbu-mcp-server.git
cd uzcbu-mcp-server
python -m venv .venv
source .venv/bin/activate  # Windows: .venv\Scripts\activate
pip install -r requirements.txt

Quick start (for testing)

python -m uzcbu_mcp.server

The server runs via the standard MCP stdio transport and waits for requests from an MCP client.

Connecting to Claude Desktop

Open the Claude Desktop configuration file (claude_desktop_config.json) and add the following:

{
  "mcpServers": {
    "uzcbu-currency": {
      "command": "python",
      "args": ["-m", "uzcbu_mcp.server"],
      "cwd": "/to'liq/yo'l/uzcbu-mcp-server"
    }
  }
}

Restart Claude Desktop. Now if you ask Claude "what is the dollar rate today?" or "convert 100 dollars to euros," it will automatically call the tools on your server.

Testing

To test the logic without going online (with mock data):

python test_offline.py

Project structure

uzcbu-mcp-server/
├── uzcbu_mcp/
│   ├── __init__.py
│   └── server.py          # MCP server va barcha tool'lar
├── test_offline.py        # Mock ma'lumot bilan offline testlar
├── requirements.txt
├── pyproject.toml
└── README.md

Why this project stands out

This is not just an "API wrapper" — it is built on the Model Context Protocol, the new standard of the AI ecosystem, meaning any MCP-compatible AI client (not just a single chatbot) can use it. MCP servers that work with Uzbekistan-specific data (official Central Bank rates) are currently very rare.

Extension ideas

• Add other CBU statistical data (inflation, interest rates)

• Speed up by caching results (for a few minutes)

• Integrate with a Telegram bot or N8N workflow

• Add HTTP/SSE transport to make the server work remotely

License

MIT