Port22

Product Introduction

1. Definition: Port22 is a macOS and iOS application suite that functions as a remote monitoring and control interface for AI-powered coding agents (LLMs) running in a local terminal. It is a productivity tool in the developer tooling and AI-assisted coding (AI pair programming) category.
2. Core Value Proposition: Port22 exists to unblock AI coding agents by providing remote approval and real-time monitoring from an iPhone. It solves the critical workflow interruption where a developer must be physically present at their Mac to approve file edits or commands, enabling asynchronous agent operation and increased productivity.

Main Features

1. Non-Invasive Session Attachment: Port22 does not launch or wrap the AI agent process. Instead, its lightweight Mac agent binary attaches to existing terminal sessions (TTYs) running compatible CLI tools like Claude Code, Cursor.sh's Codex, or OpenCode. It monitors the session's input/output stream in real-time.
2. Real-Time, Bi-Directional Stream to iOS: The core technology is a low-latency, live data stream between the Mac agent and the paired iPhone app. This stream transmits every token of the agent's output (thoughts, code diffs, shell commands) and allows the user to send approval decisions (e.g., y/n/a) back to the terminal session from their phone.
3. Hybrid LAN & Encrypted Relay Connectivity: The connection is LAN-first for optimal performance and privacy on the same network. When off-network, it automatically fails over to an end-to-end encrypted relay. The relay forwards encrypted payloads but cannot decrypt them, as the encryption keys are only on the user's Mac and iPhone.
4. Apple ID-Based Zero-Config Pairing: Leveraging the native Apple ecosystem, devices signed into the same iCloud account discover each other automatically, eliminating manual IP configuration, QR codes, or separate Port22 accounts. For enterprise scenarios with different Apple IDs, a fallback 6-character code enables local Wi-Fi pairing.
5. Context-Aware iOS System Integration: The iPhone app provides live activity updates in the Dynamic Island and on the Lock Screen, showing active agent status. It sends actionable push notifications the moment an agent requires input or completes a task, allowing immediate response.

Problems Solved

1. Pain Point: The "Agent Blockage" problem. Developers using CLI-based AI coding agents cannot leave long-running tasks unattended because the agent will pause indefinitely, waiting for human approval for file writes or shell commands, leading to wasted time and context switching.
2. Target Audience: The primary user persona is the professional software engineer or technical lead who actively uses AI coding assistants like Claude Code or Cursor in their terminal workflow. Secondary users include tech-savvy founders, indie hackers, and DevOps engineers who automate coding tasks and need oversight without being desk-bound.
3. Use Cases:
   • Running lengthy refactors or feature implementations with an AI agent, then moving to a meeting or leaving the desk while monitoring progress on a phone.
   • Approving sensitive file system or shell operations from a mobile device when away from the primary workstation.
   • Monitoring multiple concurrent AI agent sessions across different projects or machines from a single, centralized mobile interface.

Unique Advantages

1. Differentiation: Unlike IDE plugins or terminal emulators with built-in AI, Port22 is agent-agnostic and terminal-agnostic. It works with the user's existing toolchain (Claude Code, Cursor's Codex, OpenCode) in their preferred terminal (kitty, WezTerm, tmux), requiring no change in primary workflow. It is not a trial-limited freemium model; its free tier offers full functionality with reasonable caps.
2. Key Innovation: The "attach, don't wrap" architecture is the fundamental technical innovation. By interfacing directly at the TTY level, it avoids the complexity, instability, and configuration overhead of process forking or API wrapping. This, combined with native Apple ecosystem integration for seamless discovery, creates a frictionless user experience unmatched by generic remote desktop or VNC solutions.

Frequently Asked Questions (FAQ)

1. Does Port22 work with Cursor AI or GitHub Copilot? Port22 works with AI agents run via their Command Line Interface (CLI). It is compatible with Cursor's codex CLI agent and Claude Code. It does not directly interface with IDE-integrated versions of GitHub Copilot or Cursor's GUI interface unless they are operating via a supported CLI in a terminal session.
2. Is my source code secure with Port22's relay service? Yes. All data transmitted off your local network uses end-to-end encryption (E2EE). The relay server acts only as a blinded packet forwarder for encrypted data; it never possesses the decryption keys and therefore cannot read your code, agent transcripts, or commands.
3. Can I use Port22 with a work Mac that has a different Apple ID? Yes. If you cannot use the same Apple ID (e.g., on a corporate-managed Mac), you can initiate pairing directly. The Mac agent will display a 6-character code, which you enter in the iPhone app to establish a direct connection over your local Wi-Fi network.
4. What are the limitations of the free tier versus the unlimited plan? The free tier is fully functional but caps usage at one paired Mac and two simultaneous active agent sessions. The Unlimited plan removes these caps, allowing you to monitor and control an unlimited number of Macs and concurrent sessions, which is essential for developers managing multiple machines or complex multi-agent workflows.
5. How does Port22 affect the performance of my AI coding agent? The Mac agent is designed to be lightweight and non-intrusive. It operates by reading from and writing to the terminal's pseudoterminal (pty), which adds negligible overhead. It does not intercept or process the LLM's API calls, so it does not impact the core reasoning or generation speed of your AI agent.