Sayori

Name: Sayori
Author: VBN Technologies

Android app by VBN Technologies. Productivity · VBN Technologies

Store rating: Unknown
Store rating count: Unknown
Download price: Free to download
In-app purchases: Not listed in captured metadata
Version: 1.1.0
Listing last refreshed: 2026-09-11

Store description excerpt

Sayori is a private, on-device voice assistant for Android that you point at your own AI. Bring an API key for a hosted model, or point it at a local server on your own network — Sayori never bundles or requires a specific provider, and it has no servers of its own. Speech recognition, wake-word detection, and speaker verification all run entirely on your phone. Your microphone audio never leaves the device. Unlike a typical voice assistant, Sayori is also an agent: it speaks the open Model Context Protocol (MCP), so it can use real tools, and it can act on your phone directly — placing calls, opening apps, drafting texts, and more — always with your explicit confirmation first. KEY FEATURES • Voice chat — talk to your AI using on-device Whisper speech-to-text, with streamed replies read back via the system text-to-speech engine. • "Hey Sayori" wake phrase — fully offline wake-word detection, no always-listening cloud service. • Fast, natural turn-taking — an on-device neural voice-activity detector ends your turn about a second after you stop talking, instead of the multi-second pause typical of older assistants. • Speaker verification — enroll your voice so wake-word activation (and, optionally, ongoing conversation) responds only to you, not to other people or the TV in the background. • Bring your own AI — works with OpenAI, native Anthropic Messages API, Ollama, LM Studio, vLLM, LocalAI, and any other OpenAI-compatible endpoint you configure. • MCP tool support — connect Model Context Protocol servers to give your AI real capabilities: search, code execution, file access, and more. • Device tools — let the AI open apps, place phone calls, draft text messages, look up contacts, and check your location. Every single invocation requires an explicit on-screen confirmation before anything happens — texts are never sent without you personally pressing send. • Multiple persistent chat sessions, JSON export/import, and token-by-token streaming with cancellation. PRIVACY, BY ARCHITECTURE • No telemetry. No analytics. No ads. No accounts. • Microphone audio is processed entirely on-device — wake word, speech-to-text, and speaker verification never touch a network. • Chat text is sent only to the LLM endpoint URL you personally configure. Sayori has no backend of its own to collect anything. • Your API key is stored in Android's hardware-backed encrypted Keystore, never in plain text. • Full details in the in-app Privacy Policy. PERMISSIONS, ONLY WHEN YOU USE THEM Every permission is tied to a feature you explicitly enable. Microphone powers wake-word detection and voice input. Phone, contacts, and location are used only by the optional device-tool feature, only when the AI proposes an action, and only after you confirm it on-screen — Sayori does not hold SMS-sending permission at all; drafting a text opens your own messaging app with the message pre-filled for you to review and send. Full rationale for every permission is in the in-app About sc

Review coverage

AppRill has captured 0 reviews for this app. This is a collected sample, separate from the store rating count. Latest review capture: Not captured yet.

Recent captured chart positions

No chart positions have been captured for this app yet.

Chart position, rating count and review coverage do not establish downloads, revenue or profit. Understand the differences.