PyGPT

Open-source desktop AI assistant with agents, RAG and local models

Advanced
Visit PyGPTBack to AI tools

Own PyGPT? Verify ownership to get a dofollow link

Screenshot of PyGPT, Open-source desktop AI assistant with agents, RAG and local models

What is PyGPT?

PyGPT is an AI chat bot tool. Free and open-source desktop AI assistant for Windows, macOS and Linux. Connects cloud and local Ollama models, adds agents, LlamaIndex RAG, Python and MCP tools, voice, vision and image or video generation, all driven by your own API credentials.

PyGPT is a Python-based desktop application that turns a chat window into a full local AI workspace for Windows, macOS and Linux. It connects to cloud providers such as OpenAI, Google Gemini, Anthropic Claude, xAI Grok, DeepSeek, Perplexity and HuggingFace, and to OpenAI-compatible endpoints, while also running Ollama models like Llama, Qwen, Mistral, Gemma and gpt-oss entirely on the machine. Because it uses your own provider credentials, billing stays with the model vendor and local work needs no external API key at all. The application moves past single-turn chat through agents, Agent Skills and LlamaIndex RAG. Primary, orchestrator and swarm workflows coordinate multiple agents, portable SKILL.md files extend what they can do, and project-level AGENTS.md rules keep behaviour consistent inside a repository. Retrieval indexes files, webpages, media, external data and conversation history into built-in vector stores, with separate indexes per project. Work modes cover Chat, Research, Completion, Realtime audio, image and video generation, Computer use, Experts and autonomous operation, so one interface handles a wide range of tasks. Automation is where the product separates from simpler chat clients. Built-in Python, IPython and system tools run with sandbox and Docker options, MCP connectors and provider-side tools reach outside the app, and a crontab-style scheduler fires prompts on a timetable. Plugins cover files, web search, Google services, GitHub, Slack, Telegram and X/Twitter. Around the chat live a Notepad, Painter sketch tool, Terminal, media tools and an agent workflow monitor, plus voice input, speech synthesis and realtime voice conversations. PyGPT suits developers and technical users who want model choice, local execution and scriptable control from a native app. Non-technical users may find the configuration surface large. It is free and open source, available as an MSI installer, ZIP archive, Snap, AppImage, Microsoft Store release or PyPI package, with source on GitHub.

How do you use PyGPT?

  1. 1Download and install PyGPT
    Pick your platform from the download section. Windows users can grab the MSI installer or the Microsoft Store listing, Linux users have ZIP, Snap and AppImage options, and macOS users install through PyPI with pip install pygpt-net.
  2. 2Connect a model provider
    Open the settings and enter API credentials for a cloud provider such as OpenAI, Gemini or Claude, or point PyGPT at a local Ollama instance. Custom and OpenAI-compatible endpoints can be added the same way.
  3. 3Choose a work mode and preset
    Select Chat for conversation, Agents for multi-agent workflows, Research for retrieval-heavy tasks, or Realtime for voice. Presets bundle model, prompt, RAG and tool settings so a configuration can be reused.
  4. 4Index your material with RAG
    Add files, webpages or conversation history as data sources and let the built-in LlamaIndex indexing place them into a vector store. Separate indexes can be kept per project for cleaner retrieval.
  5. 5Extend with tools, plugins and MCP
    Enable Python and system tools, connect MCP connectors, or install plugins for Google services, GitHub, Slack and other targets. Sandbox and Docker options control how shell and Python code execute.
  6. 6Automate recurring tasks
    Use the scheduler to run prompts on a timetable, for example a morning summary or a periodic web check. Scheduled jobs reuse existing presets and run while the desktop app is active.

Pros and cons

Pros

  • Free and open source with the full codebase available on GitHubAI
  • Mixes cloud providers with local Ollama models, so private work can stay offlineAI
  • Agent workflows, Agent Skills and LlamaIndex RAG are built in rather than bolted onAI
  • Python, IPython, system tools and MCP connectors support real automationAI
  • Ships for Windows, macOS and Linux, including MSI, Snap, AppImage and PyPI routesAI

Cons

  • Requires your own provider API keys, so model costs are billed outside the appAI
  • The feature set is dense, and a first-time setup means many settings and panelsAI
  • No built-in hosted service: everything runs on your own machine, so you manage installs and updatesAI
  • Documentation is split between the site and Read the Docs, with no unified help centreAI
  • Local model use still depends on your hardware and Ollama being set up separatelyAI

How much does PyGPT cost?

Pricing

PyGPT is free and open source. There is no paid tier on the site; the project accepts donations through Buy Me a Coffee, PayPal and GitHub Sponsors. Model usage is paid to whichever provider you connect, while local Ollama models run without external API keys.

Learn more

Support

Installation and usage documentation is hosted on Read the Docs, and the project offers donation-based support channels. No dedicated help desk or ticket system is described on the homepage.

Integrations

MCP connectors, provider-side tools, OpenAI-compatible APIs and Ollama. Plugins cover Google services, GitHub, Slack, Telegram, X/Twitter and web search through DuckDuckGo, Google and Bing.

Features

Multiple cloud and local model providers, work modes including Chat, Agents, Research, Realtime, Image and Video generation, Computer use and Experts. Multi-agent primary, orchestrator and swarm workflows, Agent Skills and AGENTS.md rules, LlamaIndex RAG with vector stores, Python and OS tools with sandbox and Docker, MCP connectors, vision input, speech recognition and synthesis, a scheduler, plugins, profiles and built-in Notepad, Painter and Terminal tools.

Frequently asked questions about PyGPT

  • How much does PyGPT cost?
    PyGPT is free and open source. There is no paid tier on the site; the project accepts donations through Buy Me a Coffee, PayPal and GitHub Sponsors. Model usage is paid to whichever provider you connect, while local Ollama models run without external API keys.
  • How do you use PyGPT?
    The walkthrough on this page covers 6 steps: 1. Download and install PyGPT 2. Connect a model provider 3. Choose a work mode and preset 4. Index your material with RAG 5. Extend with tools, plugins and MCP 6. Automate recurring tasks.
  • What platforms does PyGPT support?
    PyGPT is available on Windows, MacOS and Linux.
  • What does PyGPT integrate with?
    MCP connectors, provider-side tools, OpenAI-compatible APIs and Ollama. Plugins cover Google services, GitHub, Slack, Telegram, X/Twitter and web search through DuckDuckGo, Google and Bing.
  • What are the limitations of PyGPT?
    Requires your own provider API keys, so model costs are billed outside the app. The feature set is dense, and a first-time setup means many settings and panels. No built-in hosted service: everything runs on your own machine, so you manage installs and updates.

Status

StatusActive
Views0
Outbound clicks0
Added10/21/2026

Platforms

WindowsMacOSLinux

Pricing

Free

Categories

AI Chat BotChat With ImagesImage ReaderLLM AggregatorRealtime Image GeneratorRoleplay With ImagesChatImage GeneratorProductivity