# Invoice MCP Server

A standalone Model Context Protocol (MCP) Server for the Invoice Generator application. This server bridges an AI Client (like Claude, Cursor, ChatGPT) with an existing Express.js/Supabase backend, allowing the AI to authenticate and interact with invoices.

## Features
- **login**: Authenticates against the backend and stores the JWT securely in-memory.
- **create_invoice**: Searches for a client by name and creates a new invoice for them.
- **download_invoice**: Retrieves the download link/stream for a specific invoice.

## Setup and Installation

1. **Install dependencies:**
   ```bash
   npm install
   ```

2. **Environment Variables:**
   Ensure you have a `.env` file at the root of the project with the following variable:
   ```env
   API_BASE_URL=https://suriyainvoicegenerator.in
   ```
   Replace the URL with your actual backend URL.

## Running Locally

To run the server locally for testing:

```bash
npm start
```
Note: The server uses stdio for MCP communication, so it will wait for stdin messages rather than acting as an HTTP server.

## Registering the MCP Server

You can register this MCP server in your AI client. For example, in **Cursor**, go to Settings -> Features -> MCP, and add a new server with the following command:

```bash
node /absolute/path/to/invoice-mcp-server/src/index.js
```
*(Make sure to use the absolute path to your `src/index.js` file.)*

## Architecture & Folder Structure

```
mcp-server/
├── src/
│   ├── index.js                  # Entry point, sets up stdio transport
│   ├── server.js                 # Registers MCP server and routes tool requests
│   ├── config/
│   │   └── env.js                # Environment variable configuration (dotenv & Zod)
│   ├── services/
│   │   └── api.js                # Axios client and in-memory session manager
│   ├── tools/
│   │   ├── login.js              # Tool for backend authentication
│   │   ├── createInvoice.js      # Tool for creating an invoice
│   │   └── downloadInvoice.js    # Tool for downloading an invoice
│   ├── middleware/
│   │   └── errorHandler.js       # Centralized error handler returning MCP errors
│   └── utils/
│       ├── logger.js             # Basic logger (hides sensitive data)
│       └── response.js           # MCP response formatting utilities
├── package.json
├── .env
└── README.md
```

## Sample Prompts

Once registered with your AI client, try prompts like:

- *"Login to the invoice generator with email user@example.com and password mypassword."*
- *"Create an invoice for Acme Corp with 2 Laptops at 45000 each. The invoice date is today and due in 30 days."*
- *"Download invoice number 12345 and give me the link."*