Jarvis is an open-source, privacy-first AI assistant for the desktop with 2,100+ stars on GitHub: voice and text chat, personal memory, dictation and MCP tools, all running locally on your own machine with models you choose.
Most AI assistants ship your conversations to someone else’s cloud. Jarvis inverts that: it runs on macOS, Windows and Linux, talks to a local Ollama instance or any OpenAI-compatible server you point it at, and keeps its memory of you on your device. You can speak to it and hear it work through dictation and voice chat, or type as usual, and because it speaks the Model Context Protocol it can use MCP tools to search and act beyond simple chat. The personal-memory layer means it can remember context across sessions — preferences, facts and past conversations — without that profile living on a vendor’s servers. Health and nutrition are among the domains its tooling touches, reflecting its positioning as a personal assistant rather than a coding tool. The project is under very active development, with updates landing as recently as October 2026, which makes it one of the livelier entries in the local-assistant space.
Jarvis is free and open source. You supply the models — a local Ollama setup costs nothing beyond your hardware, while a cloud endpoint is billed by that provider, not by Jarvis.
Privacy-conscious users who want a Jarvis-style desktop assistant with voice, memory and tools, but refuse to route their personal data through a third-party cloud to get it.