Sudo — the docs.

Sudo is a local voice home assistant that runs on a Raspberry Pi. It chains openWakeWord (listening for your wake phrase), Whisper (speech-to-text), the Claude CLI (reasoning), the Home Assistant MCP server (device control), and Piper (text-to-speech) — all in Docker on one host.

Install

On a fresh Raspberry Pi OS Lite (64-bit), log in and paste:

Install on a Pibash
$curl -fsSL https://sudo.abhishekslab.xyz/install.sh | bash

In 5–10 minutes the Pi is running the onboarding at http://localhost:3000. Everything else — Docker, the voice stack, Home Assistant, MCP — is installed and configured from there, without a terminal.

Where to go next

  • Install on a Pi — what the one-liner does, prerequisites, and self-hosting the installer.
  • Onboarding flow — every stage of the single-page wizard and what happens under the hood.
  • Backup & migrate — exporting and importing .sudo-bundle archives to move between devices.
  • Upgrades — the release model and how auto-rollback works when a deploy fails verification.

Design principles

  • Local by default. Wake word, STT, and TTS never leave the device. Only the recognized transcript reaches the reasoning agent.
  • Scoped blast radius. The assistant runs as a non-admin HA user with a long-lived token limited to entities you expose.
  • No-click setup. No hand-editing YAML, no clicking through Home Assistant onboarding — the wizard drives both via API.
  • Deployment is verification. A release is considered installed only after the end-to-end smoke test passes. Otherwise, rollback.