m’kay

Product Introduction

1. Definition: m’kay is a voice interface and remote control application for macOS that enables users to interact with and manage multiple desktop-based AI coding agents (specifically Claude, ChatGPT/Codex, and Cursor) from a web browser or mobile device. It functions as a local-first orchestration layer.
2. Core Value Proposition: It exists to untether developers and technical professionals from their desks, allowing them to monitor, query, and direct their AI-powered coding sessions remotely using natural voice commands, thereby increasing productivity and workflow continuity.

Main Features

1. Local Mac Connector: A lightweight, open-source background application that runs on the user's Mac. It acts as a secure bridge, interfacing directly with the desktop applications of supported AI agents (Claude, Cursor, etc.) via system-level accessibility APIs and AppleScript. It does not stream screen data; it reads and controls the target apps programmatically.
2. Remote Voice Interface: A web-based interface accessible from any browser on a phone, tablet, or secondary computer. This interface captures voice input, transcribes it (using local or cloud-based STT), and sends the text command to the linked Mac connector. It also receives text responses from the agents and can perform text-to-speech read-back for auditory feedback.
3. Multi-Agent Awareness & Control: The core technical capability is the system's ability to identify which AI agent is currently active or has completed a task. Users can ask context-aware questions like "Which one is done?" or "What did Claude say?" and issue subsequent commands to a specific agent, all without being physically present at the Mac.
4. Flexible Deployment (Hosted or Local): Offers two architectural models. The hosted model uses the provider's web interface for convenience. The local model allows users to self-host the entire stack, including using offline speech-to-text models (like Whisper) and their own API keys for cloud services, ensuring maximum data privacy and control.

Problems Solved

1. Pain Point: Context loss and workflow interruption when stepping away from the computer during long-running AI agent tasks (e.g., code generation, refactoring). Traditional methods require being physically at the desk to check status or provide follow-up prompts.
2. Target Audience: Software developers, engineers, and technical leads who actively use AI coding assistants like Cursor, Claude Code, and GitHub Copilot (via Codex) in their daily workflow. Particularly valuable for those with fragmented schedules or who practice time-blocking.
3. Use Cases: A developer initiates a complex refactoring task in Cursor, then goes to the gym. While away, they use their phone to ask m’kay, "Is Cursor done?" Hear the summary, and then voice the next instruction: "Now, write unit tests for the refactored module." Another scenario involves reviewing code explanations from Claude during a commute via audio read-back.

Unique Advantages

1. Differentiation: Unlike universal voice assistants (Siri, Alexa) or single-agent voice plugins, m’kay is specifically engineered for multi-agent, state-aware control of professional desktop developer tools. It doesn't replace the apps; it remotely pilots the actual apps you already use, keeping all project context local to your machine.
2. Key Innovation: The "nothing is sent until you say yes" privacy model combined with local-first orchestration. The system provides a verbal read-back of the agent's response for user verification before any command is executed or data is transmitted from the remote session back to the Mac, offering a critical layer of oversight and security.

Frequently Asked Questions (FAQ)

1. How does m’kay work with AI agents on my Mac? m’kay installs a small, open-source connector app on your Mac that securely communicates with your desktop AI applications (like Cursor and Claude). You then access a web page from another device to send voice commands to this connector, which controls the apps directly.
2. Is m’kay secure for my code and data? Yes, with a strong privacy-centric design. Your code and AI sessions never leave your Mac unless you explicitly confirm a command after hearing a read-back. You can also run the entire system locally with your own APIs for complete end-to-end control.
3. What AI agents and Mac systems are compatible with m’kay? m’kay currently supports Claude, ChatGPT/Codex, and the Cursor IDE. It requires a Mac with Apple Silicon (M-series chip) running macOS 13 (Ventura) or later for optimal performance and compatibility.
4. Can I use m’kay without an internet connection on my phone? For full remote voice functionality, your remote device needs internet to load the web interface. However, if you run the local/self-hosted version, your voice processing and agent commands can be routed through your local network, keeping data off the public internet.
5. How is m’kay different from using a remote desktop app? Remote desktop apps stream your entire screen, are often laggy, and are not voice-optimized. m’kay is purpose-built for lightweight, voice-first interaction with specific AI apps, providing status checks and command execution without the overhead and security exposure of full screen sharing.