Moshi

Product Introduction

1. Definition: Moshi is an iOS terminal application leveraging the Mosh (Mobile Shell) protocol to establish direct, resilient connections between iPhones and development machines. It enables real-time interaction with AI coding agents (e.g., Claude Code) via SSH, bypassing cloud dependencies.
2. Core Value Proposition: Moshi eliminates the desk-bound requirement for AI pair-programming by providing uninterrupted terminal access during commutes, travel, or offline scenarios. Its UDP-based resilience and on-device processing empower developers to command coding agents from anywhere.

Main Features

1. Mosh-Powered Resilience:
   • How it works: Uses Mosh’s UDP protocol instead of TCP, dynamically adapting to IP changes and network switches. Sessions persist through Wi-Fi/cellular transitions, device sleep, or brief disconnections.
   • Technologies: Mosh (Mobile Shell), UDP with forward error correction, state synchronization.
2. On-Device Voice Commands:
   • How it works: Integrates OpenAI’s Whisper model for local speech-to-text conversion. Spoken commands (e.g., "git push origin main") translate directly to terminal input without cloud processing.
   • Technologies: Whisper iOS implementation, CoreML for offline inference.
3. Biometric SSH Key Security:
   • How it works: Stores SSH keys in Apple’s Secure Enclave-backed Keychain. Face ID/Touch ID decrypts keys per session, preventing app-level access.
   • Technologies: iOS Keychain Services, Secure Enclave, biometric authentication.
4. Agent Notification System:
   • How it works: Monitors background terminal processes and triggers iOS push alerts when AI agents (e.g., Claude Code) require input or complete tasks.
   • Technologies: Background app refresh, local process tracking.
5. Terminal-Optimized UI:
   • How it works: Custom keyboard with hardware-key emulation (Ctrl, Esc, arrows) and Tmux shortcuts. Supports split panes, scrollback, and session restoration.
   • Technologies: SwiftUI for adaptive layouts, custom input accessory views.

Problems Solved

1. Pain Point: Developers cannot respond to AI agent prompts (e.g., Claude Code clarification requests) when away from their workstations, halting coding workflows.
2. Target Audience:
   • Remote software engineers using AI pair-programming tools.
   • DevOps professionals managing servers via CLI during travel.
   • Open-source contributors needing terminal access in low-connectivity areas.
3. Use Cases:
   • Approving AI-generated code changes from a mobile device during commute.
   • Debugging production issues via SSH while offline.
   • Voice-controlling code deployments from a coffee shop.

Unique Advantages

1. Differentiation: Unlike standard SSH clients (e.g., Termius), Moshi prioritizes Mosh’s session resilience and AI-agent interoperability. Competitors lack integrated voice input, biometric key security, and agent-specific notifications.
2. Key Innovation: Merging on-device Whisper with Mosh creates a zero-latency, offline-ready terminal pipeline—unique in mobile development tools.

Frequently Asked Questions (FAQ)

1. Does Moshi work without internet?
   Yes, Moshi maintains active terminal sessions during offline periods via Mosh’s local echo and state prediction, syncing when connectivity resumes.
2. How secure are SSH keys in Moshi?
   Keys are stored exclusively in Apple’s Keychain, encrypted by the Secure Enclave. Biometric authentication is required per session, with no cloud backup.
3. Can Moshi connect to Windows/Linux machines?
   Absolutely. Moshi supports any SSH-accessible device (Linux, macOS, Windows WSL) with Mosh server installed.
4. Is voice command processing done in the cloud?
   No. Whisper runs entirely on-device, ensuring voice data never leaves your iPhone.
5. What happens if my phone sleeps during a task?
   Mosh’s UDP protocol preserves session state. Upon wake, Moshi automatically re-syncs with your dev machine without manual reconnection.