# MCP Prototype – Translate Server
This project implements an **MCP (Model Context Protocol) server** for text translation.
The server supports two transports:
* **stdio** – for integration with Claude Desktop
* **http (streamable)** – for testing and working via HTTP + SSE
---
## ⚙️ Requirements
* Node.js >= 18
* Yarn or npm
* Installed [Claude Desktop](https://claude.ai/download) (for stdio integration)
* Lingvanex Translator account for text translation
---
## 🔑 Lingvanex Translator Setup
To use the Lingvanex Translator you'll need a Lingvanex account.
1. If you don't have one, [sign up for free](https://lingvanex.com/account/)
2. Go to the **Cloud API** tab: [Cloud API](https://lingvanex.com/account/#b2b)
3. Fill out the **Billing Address** data
4. Click **Continue to payment**
* To get a free trial, it is **not necessary** to add your payment card
5. Your **API key** will be generated and visible in the **Cloud API** tab: [API key](https://lingvanex.com/account/#b2b)
Now you are ready to start using the translation API.
Below is a video tutorial of the overall process (if available on Lingvanex site).
---
## 🚀 Installation & Build
```bash
# Clone the repository
git clone https://github.com/you/mcp-prototype.git
cd mcp-prototype
```
# Install dependencies
```bash
yarn install
```
---
## 🔌 Run in stdio mode (Claude Desktop)
**stdio** mode is used by Claude Desktop to connect to local MCP servers.
### Set environment variable:
TRANSPORT=stdio
### Start the server:
```bash
yarn build
yarn start
```
### Expected output:
```
MCP stdio transport running
Translate MCP Server ready
```
---
## 🌐 Run in HTTP mode (streamable)
**http** mode runs a local HTTP server with HTTP transport.
Useful for browser testing or with `curl`.
### Set environment variables:
```bash
TRANSPORT=http
HTTP_PORT=3000
```
### Start the server:
```bash
yarn build
yarn start
```
### Test the server:
```bash
curl http://127.0.0.1:3000/ping
```
**Expected response:**
```json
{ "status": "ok", "transport": "http" }
```
### Use MCP Inspector for debugging:
```bash
npx @modelcontextprotocol/inspector
```
In the MCP Inspector UI, select Transport Type - Streamable HTTP; URL - http://localhost:3000/mcp. Click Connect.
---
## 🖥️ Integration with Claude Desktop
Claude Desktop discovers local MCP servers via config file:
* **Windows**: `%APPDATA%\Claude\claude_desktop_config.json`
* **macOS**: `~/Library/Application Support/Claude/claude_desktop_config.json`
* **Linux**: `~/.config/Claude/claude_desktop_config.json`
### Example config (Windows)
Open (or create) `claude_desktop_config.json` and add:
```json
{
"mcpServers": {
"translate": {
"command": "node",
"args": [
"C:\\Users\\path\\to\\project\\dist\\index.js"
]
}
}
}
```
> ⚠️ Make sure to update the path to your local `dist/index.js` after build!
---
## ✅ How to verify
1. Launch Claude Desktop.
2. Enter a request like:
*"Use the MCP tool `translate_text` to translate 'Hello world' into Russian."*
3. If everything is configured correctly, Claude will call your MCP server and return the translation.
---
## 📌 Available Tools
### `translate_text`
Translate text from one language into another.
**Arguments:**
* `text` – the text to translate
* `sourceLang` – source language code (e.g. `"en"`)
* `targetLang` – target language code (e.g. `"ru"`)
**Example request:**
```json
{
"tool": "translate_text",
"args": {
"text": "Good morning",
"sourceLang": "en",
"targetLang": "fr"
}
}
```
**Example response:**
```json
{
"content": [
{ "type": "text", "text": "Bonjour" }
]
}
```
---
