azure-ai-travel-agents

The azure-ai-travel-agents MCP suite provides multiple specialized MCP servers that enable an orchestration workflow for travel planning. Each server exposes a specific capability as a tool that can be invoked by MCP clients: Customer Query Understanding extracts user preferences from natural language inputs, Destination Recommendation suggests suitable travel options based on those preferences, Itinerary Planning generates a detailed travel plan, and Echo Ping serves as a simple round-trip MCP server for testing connectivity. By combining these servers with LangChain.js, LlamaIndex.TS, or the Microsoft Agent Framework, the application coordinates multiple agents to respond to user queries with coherent itineraries and destination suggestions. Use the MCP clients to route requests to the appropriate server and chain responses across servers to build end-to-end travel experiences.

Prerequisites:

• Docker (for local previews and the Docker Model Runner workflow)
• Node.js v22+ (for the Node.js MCP server variant)
• Python 3.9+ (for Python MCP servers)
• Java JDK 11+ (for Java-based MCP servers)
• .NET SDK (if you plan to run the .NET MCP server)

Step-by-step installation:

1. Install prerequisites on your machine according to your platform (Windows/macOS/Linux).
2. Clone the repository:

git clone https://github.com/Azure-Samples/azure-ai-travel-agents.git
cd azure-ai-travel-agents

3. Install Node.js MCP variant (Destination Recommendation):

cd path/to/destination-recommendation
npm install
# or if using pnpm/yarn, as appropriate
npm run build

4. Install Python MCP variants (Customer Query Understanding and Echo Ping):

pip install -r requirements.txt

5. Install Java MCP variant (Itinerary Planning):

# If a build is provided
mvn clean package

6. Run MCP servers (examples):

# Node.js server
node dist/server.js

# Python servers
python -m customer_query_understanding
python -m echo_ping

# Java server
java -jar itinerary-planning.jar

7. Verify MCP client connectivity and ensure the servers appear in the MCP registry. For local testing, you can use MCP clients to call each server and confirm responses.

Notes and tips:

• The repository supports multi-language MCP servers; ensure each server's dependencies are installed according to its language.
• If you are previewing locally, Docker Model Runner can help emulate the MCP environment and containerized execution.
• Review the adventure.config.json and llms.txt files to understand how LLMs are guided to utilize these MCP tools effectively.
• If you encounter port conflicts, adjust the server port mappings in the run commands or container configurations.
• Ensure environment variables for any API keys or credentials are set as needed by each server (e.g., MCP_AUTH_TOKEN, API_KEYS).