ts-travel

This MCP server implements a complete travel booking backend written in TypeScript. It exposes 8 MCP tools (e.g., search, book, cancel, and related operations) that AI clients can call via standard MCP IO flows. The server maintains mock flight data and a booking state with persistence, so changes from one AI client are visible to all others and survive across sessions. You can route AI client requests through the MCP server, then receive structured responses for flight options, seat availability, and booking confirmations. The project supports multiple AI clients (Claude, Codex, Gemini) by design, allowing simultaneous interactions with the same underlying data model.

To use it with an AI client, first build the backend, then start the MCP server. The built dist/index.js serves as the entry point for stdin/stdout based communication. When integrated with an AI client, your client will send requests to the MCP server, which executes the appropriate tool logic in database.ts and returns a formatted response. This setup enables real backend state management and consistent data across UI and CLI integrations.

Prerequisites

• Node.js v18 or higher
• npm (comes with Node.js)
• Git (optional, for cloning)

1. Clone the repository (or download the package)

git clone <repository-url> travel-mcp-server
cd travel-mcp-server

2. Install dependencies

npm install

3. Build the backend (TypeScript to JavaScript)

npm run build

4. Run the MCP server

npm start

5. (Optional) Run the API for the web app and/or CLI as described in the README's Quick Start section. For the MCP-only path, the built dist/index.js will be used by AI clients.

Prerequisite notes:

• Ensure you have network access if you plan to connect external AI clients or frontend apps.
• The build step outputs to dist/index.js and dist/database.js, which the MCP server uses at runtime.

Tips and common issues:

• If you modify TypeScript sources, remember to run npm run build to refresh dist/index.js and dist/database.js.
• The MCP server persists state for bookings and seat availability; ensure the data directory remains writable in environments with persistence.
• When integrating with AI clients, ensure the IO channel (stdin/stdout) is properly wired to the client as described in the PROJECT_WALKTHROUGH.md guides.
• If you encounter port or network issues, verify that nothing else is occupying the default ports used by the web app and API servers when running the full stack.
• Environment variables can be added under the env field in mcp_config if your deployment requires custom configuration (e.g., API keys, debug flags).