Exely Hotel Booking MCP Assistant

A project to integrate the Exely Distribution API with an LLM via an MCP server, allowing interaction through a Telegram bot.

This project is containerized with Docker for easy and reliable deployment in a production environment on any server or VPS.

Key Features

• One-Click Deployment: Use the deploy.sh (for Linux) or deploy.ps1 (for Windows) scripts for automatic installation and setup.

• Interactive Setup: The scripts will prompt you for the necessary API keys and create the configuration files for you.

• Flexible: You can deploy all components on a single machine or split the MCP server and Telegram bot across different servers.

• Production-Ready: Uses a Multistage Docker build to create lightweight and secure images.

• Reliable: Containers are configured to restart automatically in case of failure.

Project structure

exely_mcp_project/
├── app/                      # Application source code (FastAPI/MCP server)
│   ├── __init__.py
│   ├── config.py             # Application configuration (reads from .env files)
│   ├── main.py               # FastAPI entrypoint with the MCP server
│   ├── exely_client/         # Client for the Exely API
│   │   ├── __init__.py
│   │   ├── client.py         # The API client itself
│   │   └── schemas.py        # Pydantic models for the Exely API
│   ├── llm_client/           # Client for the Mistral LLM API
│   │   ├── __init__.py
│   │   └── llm_client.py
│   └── mcp_tools/            # MCP tools for the LLM
│       ├── __init__.py
│       ├── schemas_llm.py    # Pydantic models for tool parameters
│       ├── tools.py          # Tool implementation logic
│       └── prompt_utils.py   # Utilities for generating prompts
│
├── telegram_bot.py           # Telegram bot source code
├── pyproject.toml            # Project dependencies and metadata
│
├── README.md                 # Main documentation (English)
├── README.ru.md              # Optional documentation (Russian)
│
│   --- Deployment Files ---
├── deploy.sh                 # Deployment script for Linux/macOS
├── deploy.ps1                # Deployment script for Windows (PowerShell)
├── Dockerfile                # Instructions for building the Docker image
├── .dockerignore             # Specifies files to exclude from the image
├── docker-compose.yml        # Docker Compose for all-in-one deployment
├── docker-compose.server.yml # For deploying the server only
├── docker-compose.bot.yml    # For deploying the bot only
│
│   --- Environment Files (usually not in git) ---
├── .env.prod                 # (Generated) Production environment variables
├── .env.bot.prod             # (Generated) Env vars for a remote bot
└── .env.example              # Template for local development (without Docker)

Related MCP server: Amadeus Hotel API MCP Server

Deployment (Production)

This is the recommended method for running the project on a VPS or any other server.

Prerequisites

• Git to clone the repository.

• Docker and Docker Compose: The deployment script will attempt to automatically install them on Ubuntu systems. For other operating systems, please install them according to the official documentation.

Launch Instructions

1. Clone the repository to your server:

   git clone https://github.com/Bdata0/Exely_MCP.git
   cd ~/Exely_MCP

2. Run the deployment script: The script will check for Docker, prompt you for all required API keys and tokens, create the configuration files, and launch the project.

   • For Linux (Ubuntu, Debian, etc.): First, make the script executable:

     chmod +x deploy.sh

     Then, run it:

     ./deploy.sh

   • For Windows (using PowerShell): You may need to allow script execution for the current session:

     Set-ExecutionPolicy -ExecutionPolicy RemoteSigned -Scope Process

     Then, run the script:

     .\deploy.ps1

3. Follow the on-screen instructions:

   • The script will ask you to enter your EXELY_API_KEY, MISTRAL_API_KEY, and TELEGRAM_BOT_TOKEN. Input will be hidden for security.

   • Next, it will prompt you to choose a deployment scenario:

     1. All-in-one: Run both the server and the bot on the current machine (most common choice).

     2. Server only: Run only the MCP server.

     3. Bot only: Run only the Telegram bot (will require the IP address of the server machine).

After selecting a scenario, the script will automatically build the Docker images and start the containers in the background.

Application Management

• Check container status: docker compose ps

• View real-time logs: docker compose logs -f

• Stop the application: docker compose down

Your bot is now fully configured and running!