🧠 Advanced MCP Server Setup with uv, llama-index, ollama, and Cursor IDE

✅ Prerequisites

• Python 3.10+ installed
• uv (by Astral) installed globally (pip install uv)
• Ollama installed and running locally
• Cursor IDE installed

🛠 Step 1: Project Setup

1.1 Create a New Project Directory

uv init mcp-server
cd mcp-server

1.2 Create and Activate Virtual Environment

uv venv
.venv\Scripts\activate  # On Windows
# OR
source .venv/bin/activate  # On Linux/Mac

🔐 Step 2: Environment Configuration

Create a .env file in the root of your project and add your API key:

LINKUP_API_KEY=your_api_key_here

📦 Step 3: Install Required Dependencies

Run these commands one by one inside your virtual environment:

# Core MCP CLI and HTTP utilities
uv add mcp[cli] httpx

# Linkup SDK for orchestrating agents
uv add linkup-sdk

# LlamaIndex integrations
uv add llama-index
uv add llama-index-embeddings-huggingface
uv add llama-index-llms-ollama

# Optional: for using notebooks
uv add ipykernel

🧪 Step 4: Confirm Installation

After installation, check your uv-managed pyproject.toml for something like this:

[tool.uv.dependencies]
mcp = { extras = ["cli"] }
httpx = "*"
linkup-sdk = "*"
llama-index = "*"
llama-index-embeddings-huggingface = "*"
llama-index-llms-ollama = "*"
ipykernel = "*"

⚙️ Step 5: Create a Minimal Server Entry Point

Create a server.py file inside the project root:

# server.py

from mcp.cli import app

if __name__ == "__main__":
    app()

You can later replace this with your own FastMCP or Agent orchestrator script.

🧠 Step 6: Run Ollama Locally

Make sure Ollama is installed and running:

ollama run llama3.2 Or any model you want

This starts the LLM backend at http://localhost:11434.

🖥️ Step 7: Configure MCP Server in Cursor IDE

7.1 Open Cursor Settings

• Open Settings → Go to MCP section.
• Click on "Add New Global MCP Server"

7.2 Fill Out the Configuration

Replace the paths with your actual machine paths. You can get the full path to uv by running:

where uv  # Windows

Now add this to your Cursor IDE settings:

{
  "mcpServers": {
    "weather": {
      "command": "C:\\Users\\SIDHYA\\AppData\\Roaming\\Python\\Python311\\Scripts\\uv.exe",  // Replace with your actual uv path
      "args": [
        "--directory",
        "C:\\Users\\SIDHYA\\Development\\Ai\\mcp-server",
        "run",
        "server.py"
      ]
    }
  }
}

🧪 Step 8: Test the Integration

1. Open any .py file in Cursor.
2. Use the MCP tools (usually accessible via ⌘K or Ctrl+K) to run the “weather” MCP server.
3. You should see the server spin up using your server.py.

📘 Suggested Directory Structure

mcp-server/
├── .env
├── pyproject.toml
├── server.py
└── rag.py

🔁 Keep Things Updated

To update dependencies:

uv pip install --upgrade llama-index
uv pip install --upgrade linkup-sdk

✍️ Author

👋 Hey, I'm Asutosh Sidhya

🌐 Connect with Me

• 📧 Email: sidhyaasutosh@gmail.com
• 🧑‍💻 GitHub: @asutosh7
• 💼 LinkedIn: linkedin.com/in/asutosh-sidhya

If you're building something around AI agents, local LLMs, or automated RAG pipelines—I'd love to connect or collaborate!