MCPOllamaStack

MCP Ollama Stack

A Model Context Protocol (MCP) chat application that integrates Ollama with a custom MCP server to provide AI-powered conversations with tool integration capabilities.

Project Structure

MCPOllamaStack/
├── MCPServer/          # MCP Server application
│   └── Program.cs      # Server implementation
└── MCPClient/          # MCP Client application  
    └── Program.cs      # Client implementation with chat interface

Overview

This project consists of two main components:

• MCP Server: A web application that exposes tools via the Model Context Protocol using stdio transport
• MCP Client: A console chat application that connects to both Ollama and the MCP server to provide an interactive AI experience with tool capabilities

Prerequisites

Required Software

1. .NET SDK (version 8.0 or later)
2. Ollama - Local AI model server

• Download from ollama.ai
• Must be running on http://localhost:11434

Required Model

The project is configured to use the llama3.2:latest model. Ensure you have this model installed:

ollama pull llama3.2:latest

Setup Instructions

1. Start Ollama

Make sure Ollama is running with the required model:

# Start Ollama (if not already running)
ollama serve

# Verify the model is available
ollama list

2. Run the Project

Navigate to the client directory and run the application:

cd MCPClient
dotnet run

The client will automatically:

• Start the MCP server from the parent directory
• Connect to Ollama at http://localhost:11434
• Initialize the chat interface
• Display available tools from the MCP server

Usage

Chat Interface

Once the application starts, you'll see:

1. A list of available tools from the MCP server
2. A prompt asking for your query

Available Commands

• Regular chat: Type any message to interact with the AI
• Clear history: Type /clear to reset the conversation history
• Exit: Press Ctrl+C to stop the application

Example Session

Tool: example-tool - Example tool description

Query:
Hello, what can you help me with?

Bot:
I'm an AI assistant that can help you with various tasks. I have access to tools from the MCP server that can extend my capabilities...

Features

• Streaming responses: Real-time AI response streaming for better user experience
• Tool integration: Automatic discovery and integration of MCP server tools
• Conversation history: Maintains chat context throughout the session
• Error handling: Robust error handling for missing directories and connection issues

Architecture

The application uses:

• OllamaSharp: For connecting to the Ollama API
• Microsoft.Extensions.AI: For AI chat client abstraction
• ModelContextProtocol.Client: For MCP server communication
• Stdio transport: For communication between client and server

Troubleshooting

Common Issues

1. "MCPServer directory not found"

• Ensure the project structure matches the expected layout
• Verify you're running from the correct directory

1. Ollama connection failed

• Confirm Ollama is running: ollama list
• Check if the service is accessible at http://localhost:11434

1. Model not found

• Pull the required model: ollama pull llama3.2:latest

1. Tools not loading

• Check MCP server logs for any startup errors
• Ensure the server has proper tool implementations

Logs and Debugging

The MCP server includes console logging with trace-level output directed to stderr for debugging purposes.

Development

Adding New Tools

To add new tools to the MCP server:

1. Create tool classes in the MCPServer project
2. Ensure they're discoverable via WithToolsFromAssembly()
3. Restart the client to reload tools

Modifying the Model

To use a different Ollama model, update the model name in the client's Program.cs:

new OllamaApiClient(new Uri("http://localhost:11434/"), "your-model-name")

License

This project is provided as-is for educational and development purposes.

Source & license

This open-source MCP server is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: razpinator
• Source: razpinator/MCPOllamaStack
• License: MIT

Install and usage instructions live in the source repository linked above.