Sayonic

Product Introduction

1. Definition: Sayonic is a native macOS voice assistant application that operates from a persistent, pill-shaped UI element (the "pill") on the user's screen. Technically, it is a system-level automation and AI agent orchestration tool that leverages macOS Accessibility, Screen Recording, and Automation permissions to execute user intents expressed in natural language.
2. Core Value Proposition: It exists to eliminate manual, repetitive desktop interactions by translating spoken or typed natural language commands into direct actions across the macOS environment, including application control, file search, content creation, and task delegation to integrated coding agents. Its primary value is hands-free Mac productivity and context-aware AI automation.

Main Features

1. The Context-Aware Pill Interface: A minimal, persistent UI element that docks at the screen's edge or within the MacBook notch. It provides constant, low-friction access. The pill displays real-time status, shows transcribed commands, and offers a type-in option. It leverages macOS APIs to dynamically adapt to system light/dark mode.
2. Natural Language Command Interpreter: The core engine accepts unstructured, conversational English commands (e.g., "find my September invoice," "pause the music in Chrome"). It uses a planning AI to decompose requests into executable steps. This works by analyzing the command alongside a real-time, summarized context of the active screen and frontmost application, which is sent to the selected "brain" for processing.
3. Multi-Brain AI Orchestration: Sayonic does not rely on a single AI model. Instead, it functions as a router, allowing users to select from different "brains" per task: a built-in cloud model (Sayonic Cloud), user-provided API keys (OpenAI, Anthropic, local models via Ollama), or integrated CLI coding agents (Claude Code, Cursor CLI). This enables cost control and leverages specialized agents for complex tasks.
4. Cross-Application Action Engine: Using macOS Automation and Accessibility frameworks, it executes precise actions. This includes: launching/activating apps, controlling media playback in browser tabs (Chrome/Safari), typing text into focused fields, creating events in Calendar, appending notes to Apple Notes, and performing file searches via Spotlight (including content search).
5. Privacy-Centric Architecture: Data transmission is minimized and explicit. Only the voice audio and a brief screen summary are sent to the chosen AI brain for the active request. File contents are never sent for search; only filenames and metadata are used. API keys are stored locally in the macOS Keychain. The application does not collect analytics or retain audio recordings.

Problems Solved

1. Pain Point: Context-switching friction and manual workflow interruption. Users waste time manually navigating Finder, switching apps, searching for tabs, or formatting calendar events, breaking their deep work flow.
2. Target Audience: Knowledge workers and power users on macOS, including software developers (using integrated CLI agents), writers, researchers, project managers, and executives who perform repetitive digital tasks and value speed and automation.
3. Use Cases: A developer saying "Open my project in VS Code and run the dev server" hands off to Claude Code; a writer saying "Create a note titled 'Article Outline' and add bullet points for introduction, body, conclusion"; an admin saying "Find the PDF invoice from Acme last month and email it to John"; a user saying "Add a dentist appointment for next Friday at 2 PM to my calendar."

Unique Advantages

1. Differentiation: Unlike Siri (limited macOS integration) or global keyboard shortcut tools, Sayonic is a dedicated, always-available macOS overlay with deep system integration and multi-agent support. Unlike pure automation tools (Shortcuts, Keyboard Maestro), it requires no pre-built scripts, understanding natural language on the fly.
2. Key Innovation: Its agent-agnostic, privacy-focused orchestration layer. The "pill" is a universal front-end that can route requests to the most suitable or cost-effective backend AI (cloud, local, or coding agent), while enforcing a strict privacy model where sensitive data (file contents, calendar details) remains on-device.

Frequently Asked Questions (FAQ)

1. How does Sayonic's privacy model compare to other AI assistants? Sayonic operates on a minimal data transmission principle. For most tasks, only your spoken command and a high-level description of your active screen (e.g., "Safari window with YouTube tab") are sent to the AI. Your documents, file contents, calendar events, and note details are never uploaded for processing; local searches use macOS Spotlight.
2. Can I use Sayonic with local AI models like Llama or Mistral on my Mac? Yes, through the "API Keys" brain option. You can configure Sayonic to use an OpenAI-compatible local endpoint, such as Ollama or LM Studio. This allows for completely private, offline-capable command processing, though voice transcription may still require an internet connection depending on your setup.
3. What are "Cloud credits" and how do they differ from using my own API keys? Cloud credits are Sayonic's proprietary usage tokens for its built-in AI service. The Pro and Max plans include monthly credit allowances. Using your own keys (BYOK plan) means you pay the AI provider (e.g., OpenAI) directly, often allowing finer cost control and access to specific models not offered by Sayonic Cloud.
4. Is Sayonic a subscription service, and what happens if I stop paying? Yes, it uses a subscription model per Mac. If you cancel, the application will continue to function until the end of your current billing period, after which the premium features (including AI processing) will be disabled. The app itself may remain installed but inactive.
5. Why does Sayonic require Screen Recording and Accessibility permissions on macOS? These permissions are essential for its core functionality. Accessibility allows it to programmatically click buttons, type text, and control applications. Screen Recording allows it to analyze the content of your active window to provide context to the AI (e.g., "you are in a Chrome tab playing music"), enabling commands like "pause this."