Rainer85ah

You can deploy and use a modern MCP server foundation built with Python, FastMCP, FastAPI, Docker, Ollama, and Open-webUI to manage and expose AI/LLM capabilities. This setup is designed for rapid prototyping and production use, offering a REST API with OpenAPI docs and easy integration with local AI runtimes and a web interface.

How to use

You will run the MCP server to provide an API endpoint that lets you manage AI models and routes, use Ollama for LLM execution, and access a web-based Open-webUI chat interface. From there you can route prompts to different models, test interactions, and observe API responses in a structured, scalable way.

How to install

Prerequisites you need before installing:

• Docker and Docker Compose installed on your machine
• Git available to clone the project
• A modern Python environment if you plan to run without Docker (optional)

# Clone the project
git clone https://github.com/rainer85ah/mcp-server.git
cd mcp-server

# Run with Docker Compose (recommended)
docker compose up --build -d

# Access the app
# Ollama: http://localhost:11434
# API Docs: http://localhost:8000/docs
# OpenAPI: http://localhost:8000/openapi.json
# MCP Server: http://localhost:8000/service/mcp/
# Open-webUI: http://localhost:3000

Configuration and use notes

This setup exposes an MCP HTTP endpoint you can use to communicate with the server and manage AI models. The designated MCP URL for client access is http://localhost:8000/service/mcp/. Use the API docs at http://localhost:8000/docs to discover available endpoints, models, and routes.

If you prefer running components locally without Docker, you would start equivalent services using your chosen tools. The essential pieces to run are the API backend (FastAPI), the MCP framework (FastMCP), and the LLM orchestration (Ollama). Ensure the ports 8000 (API), 11434 (Ollama), and 3000 (Open-webUI) are available on your machine.

Security and maintenance

Keep dependencies up to date and monitor the OpenAPI docs for any newly exposed endpoints. When deploying to production, consider securing the API with authentication and rate limiting, and use Docker networking to isolate services.

Troubleshooting tips

If the MCP server does not start as expected, check container logs for the API service and the Ollama integration. Verify that required ports are not in use by other processes and that Docker Compose completed its build step successfully.

Available tools

docker

Containerized runtime for building, deploying, and running the MCP server and its components.

ollama

LLM execution and orchestration layer that integrates with the MCP workflow.

open_web_ui

Chat-style interface for interacting with AI models via a web UI.

fastapi

High-performance API backend used to expose MCP endpoints and OpenAPI docs.

fastmcp

Model Control Protocol framework managing AI models and routes.

openapi_docs

Automatically generated API documentation for endpoints exposed by the MCP server.