Moshi is a mobile interface for AI coding agents that lets you run tools like Claude Code from your iPhone. It connects directly to your development machine using resilient Mosh connections that survive network changes.
Moshi is a mobile interface for AI coding agents, purpose-built as a terminal for iPhones and iPads rather than a shrunken desktop shell. It serves developers who run agent CLIs like Claude Code, Codex, OpenCode, or Gemini and need to monitor, direct, and interact with those agents away from their desk. Dubbed the "baby monitor for your AI agents," Moshi delivers a native phone-first workflow with resilient Mosh connections that survive network switches, app kills, and device sleeps. The core value proposition is true mobile supervision and control of AI-assisted coding, enabling you to check on progress, approve requests, and provide input from anywhere — the couch, beach, coffee shop, or bed, with features like a live kanban of agent tasks and deep tmux integration.
The concrete problem Moshi solves is the limitation of being tethered to a desktop when running AI coding agents. Developers often need to approve changes, review diffs, or provide new prompts while away from their main machine. Traditional remote shells are not optimized for mobile or for the unique workflow of agent CLIs, and without a dedicated solution, developers must either stay at their desk or risk missing important agent prompts, resulting in wasted compute time and delayed feedback. Moshi addresses this with a terminal that stays connected even after network changes, and with agent-specific features that make remote management practical. The app's persistent Mosh connection prevents lost progress, and its agent kanban gives you a live overview of what each agent is doing, eliminating the painful "what did it do while I was away?" question.
The first major feature group is Agent Kanban and tmux Integration. The site describes a "live kanban of every agent's tasks, asks, and context budget" that appears in the terminal header. This lets you see at a glance what each agent is working on, what it's asking for, and how much context it's using. Deep tmux integration provides a picker for windows, gesture-based navigation, and panels built for agent CLIs. You can quickly switch between agent sessions using the tmux picker, rearrange panes with swipes and pinch zoom, and view the agent's working tree directly from the header. This is useful because tmux is the standard for managing multiple agent sessions, and Moshi makes it tactile with swipe windows, pinch zoom, and double-tap Tab — turning a complex desktop workflow into a mobile-native experience.
The second major feature group covers Paste & Annotate Images, Diff Viewer, and Voice-to-Terminal. The site highlights the ability to "crop, scribble, drop screenshots straight into your agent prompt." This means you can take a screenshot of a bug, annotate it with your phone's markup, and paste it directly into the agent's context — all without leaving the terminal. The diff viewer shows changes in the agent's working tree from the header, allowing quick review of code modifications. Voice input is another standout: on-device speech engines like Parakeet, Whisper, Apple, or cloud dictation let you dictate prompts and commands instead of typing, crucial on a phone screen. These features together make mobile agent interaction not just possible but efficient, reducing friction for tasks like reviewing diffs and providing visual context.
Publisher
A
admin
Launch Date2026-01-19
Platformmobile
Pricingfree
Domain Authority
Ahrefs DR29/100