Unichat MCP Server

🚀 Unichat MCP Server (Python Version)

This is the Python version of the Unichat MCP Server. There is also a TypeScript version available.

This server enables requests to be sent to OpenAI, MistralAI, Anthropic, xAI, Google AI, DeepSeek, and Alibaba via the MCP protocol, using tools or predefined prompts. A vendor API key is required.

✨ Features

Tools

The server implements one tool:

• unichat: Sends requests to Unichat.
  • Accepts "messages" as a required string parameter.
  • Returns a response.

Prompts

• code_review
  • Reviews code for best practices, potential issues, and improvements.
  • Parameters:
    • code (string, required): The code to be reviewed.
• document_code
  • Generates documentation for code, including docstrings and comments.
  • Parameters:
    • code (string, required): The code to be commented on.
• explain_code
  • Explains in detail how a piece of code works.
  • Parameters:
    • code (string, required): The code to be explained.
• code_rework
  • Applies requested changes to the provided code.
  • Parameters:
    • changes (string, optional): The changes to be applied.
    • code (string, required): The code to be rewritten.

🚀 Quick Start

Installation

Claude Desktop

On MacOS: ~/Library/Application\ Support/Claude/claude_desktop_config.json On Windows: %APPDATA%/Claude/claude_desktop_config.json%

Supported Models:

The list of currently supported models that can be set as "SELECTED_UNICHAT_MODEL" can be found here. Make sure to set the relevant vendor API key as "YOUR_UNICHAT_API_KEY".

Example:

"env": {
  "UNICHAT_MODEL": "gpt-4o-mini",
  "UNICHAT_API_KEY": "YOUR_OPENAI_API_KEY"
}

Development/Unreleased Server Configuration:

"mcpServers": {
  "unichat-mcp-server": {
    "command": "uv",
    "args": [
      "--directory",
      "{{your source code local directory}}/unichat-mcp-server",
      "run",
      "unichat-mcp-server"
    ],
    "env": {
      "UNICHAT_MODEL": "SELECTED_UNICHAT_MODEL",
      "UNICHAT_API_KEY": "YOUR_UNICHAT_API_KEY"
    }
  }
}

Released Server Configuration:

"mcpServers": {
  "unichat-mcp-server": {
    "command": "uvx",
    "args": [
      "unichat-mcp-server"
    ],
    "env": {
      "UNICHAT_MODEL": "SELECTED_UNICHAT_MODEL",
      "UNICHAT_API_KEY": "YOUR_UNICHAT_API_KEY"
    }
  }
}

Installation via Smithery

To automatically install Unichat for Claude Desktop via Smithery:

npx -y @smithery/cli install unichat-mcp-server --client claude

🔧 Development

Build and Release

To prepare the distribution package:

1. Remove the old version:

rm -rf dist

2. Synchronize dependencies and update the lock file:

uv sync

3. Build the release:

uv build

This will create source and binary releases in the dist/ directory.

4. Publish to PyPI:

twine upload dist/*

🐞 Debugging

Start the debug server in the project root directory using poetry:

poetry run uvicorn unichat.mcp_server:app --reload

Then access http://localhost:8000 in your browser.

🛠️ Debugging Tools

Start the debug server in the project root directory using poetry:

poetry run uvicorn unichat.mcp_server:app --reload

Then access http://localhost:8000 in your browser.