PyGPT is an open-source desktop AI assistant supporting OpenAI GPT-6/GPT-5/GPT-4, Google Gemini, Anthropic Claude, xAI Grok, DeepSeek, Perplexity / Sonar, HuggingFace, OpenAI-compatible APIs, and local Ollama models such as DeepSeek, Qwen, gpt-oss, Gemma, Mistral, Llama, and others.

It combines chat and multi-agent workflows with RAG, Agent Skills, plugins and MCP, Python/system tools, vision, image and video generation, Computer use, web access, file handling, voice, automation, and extensible tools.

Features

• Desktop AI assistant for Linux, Windows and macOS.
• Chat, Chat with Agents, Realtime + audio, Research, Completion, Image/Video generation, Computer use, Experts, and legacy Agent/Autonomous modes.
• Multi-agent workflows with primary agent, orchestrator, swarm, and worker agents.
• SKILL.md-based Agent Skills with catalog browsing and GitHub/local import.
• MCP support with built-in Connectors manager and imports from Claude, Codex, OpenClaw, Cursor, VS Code, OpenCode, MCPorter, JSON, TOML and YAML.
• Integrated LlamaIndex RAG for files, webpages, external sources, media, and conversation history.
• Vector stores, embeddings, file/context indexing, and isolated project indexes.
• Built-in Python/OS tools with Python, IPython, system commands, built-in sandbox and Docker support.
• Built-in interactive Canvas.
• Local and provider-side tools/function calls, custom commands, filesystem access, and external API integrations.
• Computer use with supported OpenAI, Google and Anthropic models.
• Vision support for images, screenshots, attachments and camera capture.
• Image generation with gpt-image, Imagen, Gemini, Nano Banana, Grok, and others.
• Video generation with Veo and Sora.
• Web search with DuckDuckGo, Google and Microsoft Bing.
• Plugins for files, Python, web search, Google services, GitHub, Slack, Telegram, X/Twitter, MCP, and more.
• File and attachment handling with native upload, local processing, RAG, summaries and project-aware workdirs.
• Conversation history, short/long-term memory, calendar, day notes and @mentions.
• Speech recognition, speech synthesis and realtime voice.
• Crontab / Task scheduler.
• Reusable presets, multiple profiles, project-specific AGENTS.md, themes, translations and accessibility features.
• Built-in Notepad, Painter, Indexer, browser, text editor, Python/OS and Agent Workflow monitor.
• Fully configurable and extensible with custom models, providers, plugins, vector stores, loaders and agents.
• Uses your own provider API credentials; local models can run without external API keys.
• Free and open source.

Project Website: https://pygpt.net

GitHub: https://github.com/szczyglis-dev/py-gpt

PyPI: https://pypi.org/project/pygpt-net

Microsoft Store: https://apps.microsoft.com/detail/XP99R4MX3X65VQ

Documentation: https://pygpt.readthedocs.io/en/latest

Changelog

2.8.31 (2026-09-25)

• Refactored and improved tab management.
• Refactored and improved message rendering and event flow.
• Improved the Canvas plugin.
• Added the ability to send sketches and images directly from Painter to the model in real time via the Canvas plugin.
• Improved and fixed runtime attachment and image handling.
• Optimized memory usage and Python module initialization.
• Added a new Jev / System One plugin - see the documentation for more information.