Talk or type
Use realtime speech, local speech, or text with a companion that shows when it is listening, thinking, and working.
Desktop AI companion
A friendly face for serious local tools.
An Electron desktop companion for realtime voice, local models, visual artifacts, notes, images, search, and explicit opt-in computer control.
Talk naturally, keep routine work local, and see the notes, artifacts, images, and actions behind the conversation.
In developmentWindows & macOS
Mochi is in local desktop development.
Choose the right mode
From a quick thought to visible work.
Stay local when it fits
Route routine work through installed local models and local speech tools without requiring an API key.
Make artifacts
Turn the conversation into notes, diagrams, records, generated images, and a persistent thumbnail board.
Act only when invited
Enable computer-use mode explicitly for named app, keyboard, pointer, screenshot, and UI Automation tools.
Permission matters
Computer control stays opt-in.
Mochi keeps privileged desktop tools behind an explicit mode and validates renderer requests at the desktop boundary.
No silent computer-control mode
No bundled cloud account
No promise that every local model fits every machine
No reason to commit local keys or generated data
How it feels
Warm, capable, and visible.
Realtime voice
Local model mode
Visual artifacts
Persistent creative board
Explicit desktop tools
Product principles
The companion should show its work.
Mochi combines a friendly conversational surface with visible artifacts, diagnostics, provider choices, and local runtime controls.
Optional GPT, image, search, wake-word, and computer-control features require deliberate local configuration and may involve third-party services or operating-system permissions.
- Realtime and local voice
- Ollama-compatible models
- Artifact panel
- Image generation
- Notes and records
- Opt-in computer control
More information