A desktop assistant that talks, codes and acts on your computer, and asks before anything that matters.
Speak to it or type. It opens your apps, has coding agents build in a copy of your project, makes pages, slides and sheets, and keeps its memory in your own notes. Free and open source.
For Windows and macOS. One installer, nothing else to install. Not code-signed yet, so your system will warn once: the download page says what to click.
What the app looks like
What it does
Each of these is explained in full, with what it needs, on the Features page.
Talk to it
Voice chat that answers out loud and can be interrupted, and on Windows a wake word: "Hey JARVIS".
Does things on your PC
Opens apps and files, presses buttons by name, sets timers. Sending, deleting and paying always wait for your yes.
Coding with agents
Claude Code and Codex work in a copy of your project. Merged locally only when the project's checks pass.
Security check and fix
Finds the usual holes in a project, fixes them in a copy, shows you, and can undo.
Routines
Something you want done again and again, at the times you choose, with the same approvals as a typed message.
Artifacts
Pages, slides, sheets, charts and whiteboards, kept in one library with every version.
The Court of JARVIS
Put one idea on trial: a prosecution, analysts with web sources, a ruling worked out in code.
Your notes as its memory
Chats and memory are Markdown notes in an Obsidian vault that you can read and edit.
Private by design
Runs only on your computer at 127.0.0.1. Keys stay in the system key store. Private notes never go to a model.
Bring your own accounts
The app is free and has no account of its own. The AI work runs on the accounts and subscriptions you already have.
Claude Code
The assistant, coding jobs, the planner and the Court use Anthropic's command-line tool, signed in with your Claude subscription.
Codex
OpenAI's command-line agent, if you have it, takes coding jobs too. Jobs move to the other agent when one hits its usage limit.
Groq
A free Groq key turns your speech into text with Whisper. An OpenAI key can be used instead.
ElevenLabs
Optional. With your key, replies are spoken in an ElevenLabs voice; without it, the voices built into the app's browser engine are used.
Platforms
Windows: everything. macOS: the app with the wake word, an icon in the menu bar, shortcuts that work from any app, dictation into other apps, working in other apps' windows, step-by-step control and telling the Mac's own sound from your voice. The built-in browser panel and the small listening window are Windows-only for now. Every change is run by automated tests and, on a real Mac at GitHub, by a check that opens TextEdit, types, presses and dictates in it, takes a picture of the screen, presses each shortcut from another app, chooses from the menu-bar menu, plays "Hey Jarvis" to the wake word and measures a sound the Mac plays. Its owner has installed it on a real Mac and opens it from its icon; nobody has used it day to day yet.
Try it on your own computer
Install it, or run it from source with Python 3.11 or newer and nothing to pip install.