Qube

Qube is a desktop AI assistant for people who want to chat with models and work with private documents on their own hardware. It combines a document library with persistent memory and research tools that provide citations you can inspect.

• Document answers: Search PDFs, EPUBs and text, ask questions about individual files, and inspect the evidence retrieved for a reply.
• Editable memory: Keep context across conversations, review or change retained facts, and export or delete memories; an exclusion list prevents deleted facts from returning.
• Research with citations: Request reports using PubMed, arXiv, SEC filings and U.S. case law, or search through DuckDuckGo, Wikipedia or your own SearXNG instance.
• Model choice: Browse and download GGUF weights from Hugging Face, run them with the built-in engine, or connect to Ollama or LM Studio.
• Custom tools: Connect MCP servers and REST or GraphQL sources so chat and research can use your own tools and knowledge bases.

Qube runs on Windows, macOS and Linux. Local chat works offline, and chat, Library and memory data stay on your machine. Optional web and research requests contact outside sources when you ask for them. Hardware monitoring and diagnostic logs remain local, with no vendor telemetry.

CPU inference is available, with CUDA for NVIDIA GPUs, Vulkan for AMD and Intel GPUs, and Metal on macOS. Optional voice input, speech output and a floating desktop companion support hands-free conversations. Qube is open source under the MIT license and also offers paid features.