Jan is an open-source AI platform that transforms personal computers into powerful AI workstations. The application enables users to run large language models entirely offline on their own hardware, ensuring complete data privacy and sovereignty. Unlike cloud-based alternatives, Jan processes all conversations, documents, and files locally on the user's device, eliminating the need for internet connectivity once models are downloaded.

The platform supports a hybrid approach to AI deployment. Users can run open-source models like Llama, Mistral, Qwen, and Gemma directly from their desktop, or optionally connect to cloud providers including OpenAI, Anthropic, Google Gemini, and others when advanced capabilities are needed. This flexibility allows seamless switching between local privacy-focused processing and cloud-based computational power.

Jan includes an OpenAI-compatible local API server at localhost:1337, enabling integration with other applications and development tools. The platform supports Model Context Protocol integration, allowing AI models to interact with real applications including browser automation, web search, data analysis tools, and design applications through natural language commands.

The application features extensive customization options through extensions and plugins. Users can modify alignment settings, adjust moderation levels, and tailor the AI experience to specific preferences and requirements. Jan's architecture supports community-built extensions for enhanced functionality across productivity, development, and creative workflows.

Built on llama.cpp inference engine, Jan optimizes AI model performance on consumer hardware. The platform automatically detects and utilizes available GPU resources including NVIDIA, AMD, and Apple Silicon processors. Users download GGUF-format models from sources like Hugging Face, with Jan providing real-time visibility into VRAM and RAM requirements before installation.