py-gpt

PyGPT is a feature-rich desktop AI assistant that can run locally on Linux, Windows, and Mac. It integrates with OpenAI models and other SDKs to provide multi-modal capabilities such as chat, document-assisted conversations, image and video generation, web searching, code execution, file I/O, and workflow automation. The MCP integration allows you to start and manage the PyGPT server from a consistent interface, offering access to its command-line and programmatic features through the configured server entry. Users can interact with PyGPT to perform tasks like drafting text, analyzing documents, running Python code, querying web results, and orchestrating plugins and tools for automated workflows. If you are leveraging MCP tooling, you can connect to the server named in the configuration and utilize its capabilities as part of larger automation pipelines or client-side tooling.

Prerequisites:

• Python 3.10 or newer (but less than 3.14 as per the release notes)
• Internet access for initial installation and model usage (API keys may be required for OpenAI or other services)

Option A: Install via Python package (recommended for development or flexible environments)

1. Ensure Python and pip are installed and available in PATH
2. Create a new virtual environment (optional but recommended) python -m venv pygpt-env source pygpt-env/bin/activate # Linux/macOS pygpt-env\Scripts\activate # Windows
3. Install the PyGPT package from PyPI pip install pygpt-net
4. Run PyGPT as a server (MCP-compatible entry) python -m pygpt_net

Option B: Use provided binaries / AppImage (recommended for end-users)

1. Download the appropriate binary or AppImage from the official PyGPT download page
2. Make sure the file is executable (for AppImage: chmod +x path/to/PyGPT-AppImage)
3. Run the executable to start the application

Option C: Run via MCP configuration (example)

1. Ensure Python is installed and accessible
2. Create a minimal MCP config referencing the Python module as shown in mcp_config
3. Start your MCP client to connect to the PyGPT server

Notes:

• If you plan to use OpenAI models, set your OPENAI_API_KEY environment variable or provide credentials via the application’s settings. The README notes that you must have your own API key for OpenAI models.
• The project also supports alternative deployment methods; choose the approach that aligns with your environment (development, user install, or distribution).

Tips and common issues:

• If using the Python module entry (pygpt_net), ensure the module is installed in the active environment and that all required dependencies are available.
• For OpenAI-based models, ensure your API key is configured either via the OPENAI_API_KEY environment variable or through the application's UI settings.
• When running via MCP, keep the environment variables (like API keys) in the env block to ensure the MCP client can access them.
• If you encounter library compatibility issues, adjust your Python version within the supported range (>=3.10, <3.14) and re-create the virtual environment.
• For macOS users without binaries, consider installing from PyPI and running via Python to mimic the same behavior as the desktop app.