Ollama MCP Server

A Model Context Protocol (MCP) server that provides tools for interacting with Ollama models. This server enables AI assistants to list, chat with, generate responses from, and manage Ollama models through a standardized protocol.

🚀 Features

• Model Management: List, pull, and delete Ollama models

• Chat Interface: Multi-turn conversations with models

• Text Generation: Single-prompt text generation

• Dual Transport: Stdio (local) and HTTP (remote) support

• Railway Ready: Pre-configured for Railway deployment

• Type Safe: Full TypeScript implementation with strict typing

Related MCP server: Ollama MCP Server

📋 Prerequisites

• Node.js 18+

• Ollama installed and running locally

• For Railway deployment: Railway CLI

🛠️ Installation

Local Development

1. Clone and install dependencies:

   git clone <repository-url>
   cd ollama-mcp
   npm install

2. Build the project:

   npm run build

3. Start the server:

   npm start

Using with Cursor

Add this to your Cursor MCP configuration (~/.cursor/mcp/config.json):

{
  "mcpServers": {
    "ollama": {
      "command": "node",
      "args": ["/path/to/ollama-mcp/dist/main.js"],
      "env": {
        "OLLAMA_API_KEY": "<your-key-from-ollama.com/settings/keys>"
      }
    }
  }
}

Quick setup:

curl -sSL https://raw.githubusercontent.com/your-repo/ollama-mcp/main/config/mcp.config.json -o ~/.cursor/mcp/config.json

🏗️ Architecture

The project is structured for maximum readability and maintainability:

src/
├── main.ts                 # Main entry point
├── config/                 # Configuration management
├── server/                 # Core MCP server
├── tools/                  # MCP tool implementations
├── transports/             # Communication transports
└── ollama-client.ts        # Ollama API client

docs/                       # Comprehensive documentation
config/                     # Configuration files
scripts/                    # Deployment scripts

See ARCHITECTURE.md for detailed architecture documentation.

🔧 Configuration

Environment Variables

Cloud models

Ollama cloud models (names ending in -cloud, e.g. gpt-oss:120b-cloud) can be used in two ways:

1. Local Ollama — Keep OLLAMA_BASE_URL at http://localhost:11434 and run ollama signin so the daemon authenticates cloud requests.

2. API key — Set OLLAMA_API_KEY (from ollama.com/settings/keys). Cloud chat/generate requests are sent to ollama.com with a Bearer token automatically.

For cloud-only deployments (e.g. Railway without ollama signin), set OLLAMA_BASE_URL=https://ollama.com and OLLAMA_API_KEY as a secret.

Transport Modes

Stdio Transport (Default)

Perfect for local development and direct integration:

npm start

HTTP Transport

Ideal for remote deployment and web-based clients:

MCP_TRANSPORT=http npm start

🚀 Deployment

Railway Deployment

1. Install Railway CLI:

   npm install -g @railway/cli
   railway login

2. Deploy:

   railway up

3. Add models (optional):

   railway shell
   # Follow instructions in docs/RAILWAY_MODELS_SETUP.md

The Railway deployment automatically uses HTTP transport and exposes:

• MCP Endpoint: https://your-app.railway.app/mcp

• Health Check: https://your-app.railway.app/healthz

Docker Deployment

# Build the image
npm run docker:build

# Run locally
npm run docker:run

# Deploy to Railway
railway up

📚 Available Tools

The server provides 5 MCP tools for Ollama interaction:

1. ollama_list_models - List available models

2. ollama_chat - Multi-turn conversations

3. ollama_generate - Single-prompt generation

4. ollama_pull_model - Download models

5. ollama_delete_model - Remove models

See API.md for detailed API documentation.

🧪 Testing

Local Testing

# Test stdio transport
npm start

# Test HTTP transport
MCP_TRANSPORT=http npm start

# Test health check (HTTP mode)
curl http://localhost:8080/healthz

Model Testing

# List available models
ollama list

# Test a model
ollama run llama2 "Hello, how are you?"

📖 Documentation

• Architecture - Detailed system architecture

• API Reference - Complete API documentation

• Railway Setup - Model deployment guide

🤝 Contributing

1. Fork the repository

2. Create a feature branch

3. Make your changes

4. Add tests if applicable

5. Submit a pull request

📄 License

MIT License - see LICENSE for details.

🆘 Troubleshooting

Common Issues

"Cannot find module" errors:

npm install
npm run build

Ollama connection issues:

# Check if Ollama is running
ollama list

# Check Ollama service
ollama serve

Railway deployment issues:

# Check Railway logs
railway logs

# Verify environment variables
railway variables

Getting Help

• Check the documentation

• Review troubleshooting guide

• Open an issue on GitHub

Built with ❤️ for the AI community