MCP Servers

This repository contains two MCP (Model Context Protocol) servers:

1. File Finder MCP - for searching files
2. Whisper STT MCP - for converting speech to text

File Finder MCP Server

This is a Model Context Protocol (MCP) server that provides file search functionality. It allows you to search for files that contain a specified text fragment in their names.

Prerequisites

• Node.js (version 14 or higher)
• npm (version 6 or higher)
• Python 3.6 or higher (for HTTP server)

Installation

1. Clone or download this repository
2. Go to the project directory
3. Install dependencies:
   npm install
4. Assemble the project:
   npm run build

Starting the server

The project provides several options for launching the MCP server:

Option 1: Direct launch of MCP server

You can run the MCP server directly using Node.js:

or

This will start the server and it will listen for JSON-RPC requests on stdin/stdout.

Option 2: Launch HTTP server and MCP proxy

This option uses a Python HTTP server and an MCP proxy that forwards requests to the HTTP server:

1. First, start the HTTP server:
   npm run start:python
   or
   python main.py
2. Then in another terminal, run MCP proxy:
   npm run start:http
   or
   node build/index-http.js

Option 3: Integration with VS Code (Cline extension)

To integrate the server with VS Code and the Cline extension:

1. Find the MCP settings file:
   • Windows: %APPDATA%\Code\User\globalStorage\saoudrizwan.claude-dev\settings\cline_mcp_settings.json
   • macOS: ~/Library/Application Support/Code/User/globalStorage/saoudrizwan.claude-dev/settings/cline_mcp_settings.json
   • Linux: ~/.config/Code/User/globalStorage/saoudrizwan.claude-dev/settings/cline_mcp_settings.json
2. Add the following configuration to the mcpServers object in the settings file:

To use HTTP proxy:

Replace <ПОЛНЫЙ_ПУТЬ_К_ПРОЕКТУ> with the actual path to your project directory.

3. Restart VS Code to load the updated settings.

Available tools

MCP server provides one tool:

• search_files : Searches for files that contain the specified fragment in their names
  • Parameters:
    • fragment (string, required): Text fragment to search for in file names

Example of use

This example finds all files that contain ".py" in their names.

HTTP server (main.py)

In the root directory of the project there is a file main.py , which implements an HTTP server for searching files. This server provides a REST API for searching files that contain a specified fragment in their names.

Starting HTTP server

1. Go to the root directory of the project
2. Start the server using Python:
   python main.py
3. The server will be launched at http://localhost:8080

Using the API

To search for files, send a GET request to /search with the q query parameter:

This query will return a JSON array with information about all files that contain ".json" in their names. Each element of the array contains the following fields:

• name : file name
• path : absolute path to the file
• size : file size in bytes
• created : date and time of file creation

Example answer:

Whisper STT MCP Server

This is a Model Context Protocol (MCP) server that provides speech-to-text functionality using the faster-whisper library. It allows transcribing audio data into text with automatic language detection.

Prerequisites

• Node.js (version 14 or higher)
• npm (version 6 or higher)
• Python 3.6 or higher
• faster-whisper (install with pip install faster-whisper )

Installation

1. Clone or download this repository
2. Go to the project directory
3. Install dependencies:
   npm install pip install faster-whisper
4. Assemble the project:
   npm run build

Starting the server

The project provides several options for running the Whisper MCP server:

Option 1: Direct launch of MCP server

You can run the MCP server directly using Node.js:

or

This will start the server and it will listen for JSON-RPC requests on stdin/stdout.

Option 2: Launch HTTP server and MCP proxy

This option uses a Python HTTP server and an MCP proxy that forwards requests to the HTTP server:

1. First, start the HTTP server:
   npm run start:whisper:python
   or
   python whisper_server.py
2. Then in another terminal, run MCP proxy:
   npm run start:whisper:http
   or
   node build/whisper-index-http.js

Option 3: Integration with VS Code (Cline extension)

To integrate the server with VS Code and the Cline extension:

1. Find the MCP settings file:
   • Windows: %APPDATA%\Code\User\globalStorage\saoudrizwan.claude-dev\settings\cline_mcp_settings.json
   • macOS: ~/Library/Application Support/Code/User/globalStorage/saoudrizwan.claude-dev/settings\cline_mcp_settings.json
   • Linux: ~/.config/Code/User/globalStorage/saoudrizwan.claude-dev/settings/cline_mcp_settings.json
2. Add the following configuration to the mcpServers object in the settings file:

To use HTTP proxy:

Replace <ПОЛНЫЙ_ПУТЬ_К_ПРОЕКТУ> with the actual path to your project directory.

3. Restart VS Code to load the updated settings.

Available tools

MCP server provides one tool:

• transcribe_audio : Transcribes audio data to text using faster-whisper
  • Parameters:
    • audio_base64 (string, required): Audio data in base64 format
    • language (string, optional): Language code (e.g. "en", "ru"). If not specified, the language will be detected automatically.

Example of use

This example converts audio data to text, assuming the audio is in Russian.

HTTP server (whisper_server.py)

In the root directory of the project there is a file whisper_server.py , which implements an HTTP server for converting speech to text. This server provides a REST API for transcribing audio data into text.

Starting HTTP server

1. Go to the root directory of the project
2. Start the server using Python:
   python whisper_server.py
3. The server will be launched at http://localhost:8081

Using the API

To transcribe audio, send a POST request to /transcribe with a JSON body containing:

• audio : a base64 encoded string containing audio data
• language (optional): language code (e.g. "en", "ru")

Example request:

The answer will contain:

• text : full transcribed text
• segments : array of segments with timestamps
• language : a specific language
• language_probability : probability of detecting a language

Example answer:

Troubleshooting

• If you get the "No connection found for server" error, make sure you restart VS Code after updating your MCP settings.
• If the server does not respond, check that the path in the MCP settings is correct and points to the compiled JavaScript file.
• Make sure the server is built correctly by running npm run build before attempting to use it.
• To use an HTTP proxy, make sure that the appropriate HTTP server is running (on port 8080 for file-finder or 8081 for whisper-stt).
• If you have problems with faster-whisper, make sure the library is installed correctly and you have the necessary dependencies to work with GPU (if you are using GPU).

Project structure

Below is a list of the main project files and their purposes:

Root directory

• src/index.ts - TypeScript MCP file search server source code (direct implementation)
• src/index-http.ts - Source code for TypeScript MCP proxy for HTTP file search server
• src/whisper-index.ts - Source code for TypeScript MCP speech-to-text server (direct implementation)
• src/whisper-index-http.ts - Source code for TypeScript MCP proxy for HTTP speech-to-text server
• build/index.js - Compiled JavaScript code of MCP server for searching files
• build/index-http.js - Compiled JavaScript code of MCP proxy for searching files
• build/whisper-index.js - Compiled JavaScript code of MCP server for converting speech to text
• build/whisper-index-http.js - Compiled JavaScript code of MCP proxy for converting speech to text
• tsconfig.json - TypeScript configuration
• package.json - Description of the package and dependencies
• main.py - Python HTTP server for file retrieval
• whisper_server.py - Python HTTP server for speech to text conversion
• README.md - Project documentation (this file)