Repolode
Submit
szczyglis-dev

szczyglis-dev/py-gpt

Cross-platform desktop AI assistant supporting cloud and local models, with agents, RAG, plugins, MCP and voice.

Open-source alternative to:LM StudioChatGPT DesktopMsty

View on GitHubWebsite
Stars
2,016
Forks
353
Open issues
66
Last commit
last month

What is it?

PyGPT is an open-source desktop AI assistant written in Python for Linux, Windows and macOS, with a ChatGPT-like conversational interface. It can talk to cloud models from OpenAI, Google Gemini, Anthropic Claude, xAI Grok, Perplexity/Sonar and DeepSeek, as well as local or self-hosted models through Ollama or any OpenAI-compatible endpoint. Users supply their own provider API keys; local models need none.

Beyond plain chat, it ships work modes for Agents (Chat, Orchestrator and Swarm workflows), Realtime voice, Research, Image and video generation, Computer use and Experts. It includes LlamaIndex-based RAG indexing for documents, web pages and media, a plugin system with MCP connectors, a built-in Python/OS execution tool, web search, vision and camera input, speech synthesis and recognition, project instructions via a `.agents/` directory, and a cron-style task scheduler.

It installs from PyPI (`pip install pygpt-net`), Snap, AppImage, Microsoft Store or prebuilt 64-bit packages, and requires Python >=3.10, <3.14.

Key features

  • 01Runs as a local desktop app on Linux, Windows and macOS with a ChatGPT-style chat interface.
  • 02Connects to OpenAI, Gemini, Claude, Grok, Perplexity/Sonar and DeepSeek, plus Ollama and OpenAI-compatible endpoints.
  • 03Agents mode offers Chat, Orchestrator and Swarm multi-agent workflows with configurable worker and iteration limits.
  • 04LlamaIndex RAG indexes files, web pages and media in formats such as txt, pdf, csv, html, md, docx, json, epub, xlsx and xml.
  • 05Plugin system and MCP connectors add filesystem, Python/OS, web search, Google, GitHub, Slack, Telegram and other tools.
  • 06Realtime voice mode with OpenAI Realtime, Gemini Live and Grok, plus VAD-based automatic turn handling.
  • 07Image and video generation with models such as gpt-image, Imagen, Gemini, Nano Banana, Veo and Sora.
  • 08Computer use can drive the desktop or a Playwright browser sandbox via mouse and keyboard actions.

Use cases

  • Chatting with multiple cloud or local models from one desktop client using your own API keys.
  • Delegating multi-step coding or research tasks to Orchestrator or Swarm agent workflows.
  • Querying internal documents and PDFs through a local RAG index instead of pasting text into the prompt.
  • Running Python, IPython and system commands from the chat via the built-in tool.
  • Using voice conversations, camera input and image or video generation in a single app.

Quick start

  1. 1

    Install PyGPT from PyPI

    pip install pygpt-net
  2. 2

    Launch PyGPT

    pygpt
  3. 3

    Add API keys in Config -> Settings -> API Keys (not needed for local models)

Things to know

  • Cloud models require your own provider API keys.
  • Linux prebuilt binaries need GLIBC 2.35 or newer.
  • License is listed as NOASSERTION, so check the repository license before reuse.

In collections

Open-source AI agent frameworks & appsMCP serversRun LLMs locallyDesktop GUI apps & toolkits

Topics

#desktop-app#llm#ai-assistant#agents#rag#mcp#ollama#voice#python#cross-platform

FAQ

What is py-gpt?
Cross-platform desktop AI assistant supporting cloud and local models, with agents, RAG, plugins, MCP and voice. PyGPT is an open-source desktop AI assistant written in Python for Linux, Windows and macOS, with a ChatGPT-like conversational interface. It can talk to cloud models from OpenAI, Google Gemini, Anthropic Claude, xAI Grok, Perplexity/Sonar and DeepSeek, as well as local or self-hosted models through Ollama or any OpenAI-compatible endpoint. Users supply their own provider API keys; local models need none.
Is py-gpt free to use? What license does it have?
py-gpt is open source under the Other license. Check the license terms before using it in commercial products.
Is py-gpt still maintained?
Updated within 6 months. The last commit to py-gpt was on Oct 6, 2026.
What can I use py-gpt for?
Chatting with multiple cloud or local models from one desktop client using your own API keys. Delegating multi-step coding or research tasks to Orchestrator or Swarm agent workflows. Querying internal documents and PDFs through a local RAG index instead of pasting text into the prompt. Running Python, IPython and system commands from the chat via the built-in tool. Using voice conversations, camera input and image or video generation in a single app.
How do I install py-gpt?
Install PyGPT from PyPI: pip install pygpt-net Launch PyGPT: pygpt Add API keys in Config -> Settings -> API Keys (not needed for local models)
Where can I get py-gpt?
The source code and installation instructions are on GitHub: https://github.com/szczyglis-dev/py-gpt

Summary generated by AI from the project's README. GitHub stats refresh daily.

Related projects