PyGPT
PyGPT is an open source desktop AI Assistant powered by multiple versions of GPT, including GPT-4, GPT-4 Vision, GPT-3.5, and DALL-E 3. Compatible with Linux, W...
Last verified:
What is PyGPT?
PyGPT is an open-source, all-in-one desktop AI assistant that provides direct interaction with OpenAI language models including GPT-5, GPT-4, o1, o3, o4, and Sora2, as well as non-OpenAI models like Google Gemini, Anthropic Claude, xAI Grok, Perplexity Sonar, DeepSeek, Mistral AI, and any models provided by Ollama and LlamaIndex. The application runs locally on your desktop computer for Windows, macOS, and Linux, written in Python.
Key features include multiple work modes: Chat for natural conversation, Chat with Files for utilizing local files as additional context, Vision for image and camera-capture analysis, Agents for handling complex and autonomous tasks, Audio for audio-based interactions, Research for in-depth exploration using Perplexity and OpenAI's advanced research models, Computer Use for autonomous navigation, Completion mode, and Image and video generation. PyGPT includes a model editor and importer, executes system and custom commands, generates and runs Python code, connects to any service including web access, handles various file types, automates tasks, and enhances productivity with integrated local and external tools.
PyGPT is designed for developers, data scientists, AI enthusiasts, and anyone who wants direct access to multiple LLM providers through a single desktop interface. It supports plugin development with built-in plugins like Files I/O, Code Interpreter, Web Search, Google, Facebook, X/Twitter, Slack, Telegram, GitHub, MCP, Wikipedia, Wolfram Alpha, and many more. The application features context memory with save/load functionality, preset management for prompts, token usage tracking, voice control, accessibility features, real-time code syntax highlighting, and theme support.
PyGPT pricing
Pricing model: Free
PyGPT is completely free and open-source software under the MIT License. There is no paid tier or subscription. The application itself costs nothing to download and use. However, PyGPT uses the user's own API keys - to use OpenAI GPT models (GPT-5, GPT-4, o1, o3, etc.), you must have a registered OpenAI account with your own API key and pay for token usage directly to OpenAI. Local models via Ollama (Llama 3, Mistral, DeepSeek, Bielik, gpt-oss) do not require any API keys and are free to use. Additional API keys may be required for non-OpenAI models like Google Gemini, Anthropic Claude, xAI Grok, and Perplexity, depending on the provider's pricing. The app includes built-in token usage calculation to help users monitor their costs.
PyGPT pros
- Free and open-source under MIT License
- Supports 15+ AI models including GPT-5, GPT-4, Gemini, Claude, Grok, DeepSeek, Mistral
- 11 work modes: Chat, Chat with Files, Realtime+audio, Research, Completion, Image/video generation, Assistants, Experts, Computer use, Agents, Autonomous
- Runs locally on Windows 10/11, Linux, and macOS desktop
- Built-in LlamaIndex support for Chat with Files capability
- Vector database support with automated file embedding
- Real-time Python Code Interpreter with IPython kernel
- Speech synthesis via Microsoft Azure, Google, Eleven Labs, and OpenAI
- Speech recognition via OpenAI Whisper, Google, and Microsoft
- Web search integration with DuckDuckGo, Google, and Microsoft Bing
- Visual node-based Agents Builder for custom agent workflows
- MCP (Model Context Protocol) support with stdio, SSE, and HTTP transports
- Built-in tools: Notepad, Painter, Calendar, Indexer, Media Player, Image Viewer, Text Editor, Translator, Web Browser
- Context history with long-term memory and restoration capability
- Custom commands creation and execution functionality
- Crontab/Task scheduler included
- Supports Ollama for local models like Llama 3, Mistral, Bielik
- Real-time video camera capture in Vision mode
- Image and video generation with DALL-E 3, Imagen 3/4, and Veo 3
- Multi-language support with 18+ added languages
- Accessibility features including voice control and keyboard shortcuts
- Token usage calculation and tracking built-in
- Users own their API keys - no data transmitted to author
- Drag-and-drop file support from filesystem
PyGPT cons
- Requires user's own OpenAI API key for GPT models (costs not included)
- Binary version for Mac not available - must install from PyPi or source
- Linux version requires GLIBC >= 2.35
- Snap version requires manual connection for camera, microphone, and audio playback
- Windows version needs VC++ Redistributable installed separately
- Audio/microphone issues on Windows may require FFmpeg installation
- WebEngine/Chromium renderer may need legacy mode workaround on some systems
- Token usage calculation unavailable in Assistants API mode
- Autonomous mode can produce unexpected results when connected with plugins
- Infinite loop in autonomous mode (0 iterations) can generate high token consumption
- Computer use tool disables all other tools when selected for an expert
- Vision support unavailable in LlamaIndex Agent mode (text only)
- Streaming disabled when ReAct agent and +Tools checkbox active in Chat with Files
- Additional API keys required for non-OpenAI models (Anthropic, Google, xAI, etc.)
- Camera access in Snap version requires manual snap connect command
Frequently asked questions about PyGPT
What is PyGPT and what does it do?
PyGPT is an all-in-one desktop AI assistant that provides direct interaction with OpenAI language models (GPT-5, GPT-4, o1, o3, o4) and non-OpenAI models (Google Gemini, Anthropic Claude, xAI Grok, Perplexity Sonar, DeepSeek, Mistral AI, Ollama models like Llama 3). It runs locally on Windows, Linux, and macOS, offering multiple work modes including chat, Chat with Files, Vision, Agents, Audio, Research, Computer Use, Completion, and Image/video generation. The application includes built-in tools like Notepad, Painter, Code Interpreter, and supports over 20 plugins for extended functionality.
Is PyGPT free to use?
Yes, PyGPT is completely free and open-source under the MIT License. There are no paid plans or subscriptions for the application itself. However, to use OpenAI GPT models, you need your own OpenAI API key and must pay for token usage directly to OpenAI. Local models via Ollama (Llama 3, Mistral, DeepSeek) do not require any API keys and are free. Additional API keys may be needed for Google Gemini, Anthropic Claude, xAI Grok, and Perplexity depending on the provider.
What platforms does PyGPT support?
PyGPT supports Windows 10/11 (64-bit), Linux (64-bit with GLIBC >= 2.35), and macOS. Binary compiled versions are available for Windows (.msi) and Linux (.zip, .AppImage), but not for Mac - Mac users must install from PyPi using pip or run from source code. The application is also available via Snap Store for Linux (sudo snap install pygpt) and Microsoft Store for Windows.
What AI models can I use with PyGPT?
PyGPT supports a wide range of models including OpenAI models (GPT-5, GPT-4, o1, o3, o4, GPT-4o, gpt-image-1), Google Gemini, Anthropic Claude (Claude 4.0 Opus/Sonnet), xAI Grok (grok-4, grok-3), Perplexity Sonar models, DeepSeek (V3, R1, Coder), Mistral AI, and any model accessible through Ollama (Llama 3, Mistral, Bielik, gpt-oss, Qwen). The application includes a model editor and importer for adding custom models.
How do I set up API keys in PyGPT?
API keys are configured in the application by navigating to Config -> Settings -> API Keys. Select the provider tab (OpenAI, Anthropic, Google, xAI, Perplexity, OpenRouter, etc.), then paste your API key. For OpenAI, get your key from https://platform.openai.com/account/api-keys after registering. Local models via Ollama don't require API keys. You can configure multiple provider keys and the app will automatically use the appropriate one based on the selected model.
What work modes are available in PyGPT?
PyGPT offers 11 work modes: Chat (natural conversation with inline Vision and Image generation), Chat with Files (LlamaIndex for querying documents), Realtime + Audio (native audio input/output), Research (Perplexity API with Sonar models), Completion (text completion mode), Image and Video generation (DALL-E 3, Imagen, Veo), Assistants (OpenAI Assistants API), Experts (co-operation mode with multiple experts), Computer Use (autonomous navigation), Agent (LlamaIndex), Agent (OpenAI), and Agent (Autonomous). Each mode has specific capabilities tailored to different use cases.
How does Chat with Files mode work?
Chat with Files mode uses LlamaIndex to enable chat interaction with your documents. Files are indexed (embedded) into a vector store database by copying them to the data directory and clicking 'Index all'. Supported file types include PDF, CSV, Excel, Word, JSON, Markdown, HTML, EPUB, images, and video/audio. After indexing, you can query the entire document collection. The mode also supports querying single files on-the-fly using the query_file command. Built-in data loaders support web content from GitHub, Google Docs, YouTube, Twitter, and more.
What plugins and tools are included in PyGPT?
PyGPT includes over 20 built-in plugins: Files I/O, Code Interpreter, Web Search (DuckDuckGo/Google/Bing), Google (Gmail, Drive, Calendar, Docs, Maps), Facebook, X/Twitter, Telegram, Slack, GitHub, Bitbucket, MCP (Model Context Protocol), Wikipedia, Wolfram Alpha, OpenStreetMap, Tuya (IoT), Server (FTP/SSH), System (OS), Mouse And Keyboard, Real Time, Serial Port/USB, and more. Built-in tools include Notepad, Painter, Calendar, Indexer, Media Player, Image Viewer, Text Editor, Translator, Web Browser (Chromium), OpenAI/Google Vector Stores, Python Code Interpreter, HTML/JS Canvas, and Agents Builder (beta).
Can I use local models without API keys?
Yes, local models via Ollama do not require any API keys. You can use models like Llama 3, Mistral, DeepSeek V3/R1, Bielik, gpt-oss, and Qwen by installing Ollama and importing models through Config -> Models -> Import from Ollama. The app automatically detects Ollama and available models. For non-OpenAI cloud models like Google Gemini, Anthropic Claude, or xAI Grok, you will need their respective API keys. Local models are completely free to use once Ollama is set up.
How do I install PyGPT?
PyGPT can be installed in multiple ways: 1) Download compiled binaries from https://pygpt.net/#download for Windows (.msi) or Linux (.zip, .AppImage); 2) Install from Snap Store: sudo snap install pygpt (Linux); 3) Install from Microsoft Store (Windows); 4) Install from PyPi: pip install pygpt-net then run 'pygpt'; 5) Run from source: git clone https://github.com/szczyglis-dev/py-gpt, create virtual environment, pip install -r requirements.txt, then python3 run.py. Python 3.10-3.13 is required. Snap version requires additional connections for camera, microphone, and audio.