Remotely Respond to Claude Code and Codex Agent Prompts from iPhone

Developers often miss crucial prompts or approvals from long-running Claude Code or Codex agents when away from their Mac. This means waiting to return to the desk to provide input, interrupting workflows and causing delays in development cycles.

Pushary is a tool designed to bridge the notification gap for Codex CLI users, sending alerts to your phone when a turn finishes or a command requires approval. Unlike the native Codex Remote, which only supports the ChatGPT app, Pushary directly hooks into the CLI using Codex’s PreToolUse and PermissionRequest hooks to deliver pending commands to your mobile device for remote interaction. This allows developers to stay in control of long-running AI agents from their phones, even when the terminal UI isn’t actively monitored.

TermBridge mirrors your Mac’s terminal directly to your iPhone or iPad, providing a live, interactive view. It allows you to see agent output and type commands, giving you the necessary interface to respond to Claude Code or Codex agent prompts and approvals from anywhere.

Responding to Agent Approvals from Your Phone

The original blog post on Pushary’s blog highlights the challenge of Codex CLI users needing to approve commands. TermBridge provides the full interactive terminal, allowing you to not just receive a notification, but to actively type out your ‘yes’ or ‘no’ directly from your iPhone or iPad. This means your agent can continue its work without interruption, even if you are away from your desk. For those managing complex setups, TermBridge also helps you manage AI agent runs across multiple Macs from your iPad, ensuring all your agents stay on track.

Maintaining Control Over Long-Running AI Tasks

Long-running Claude Code or Codex agent processes often require intermittent monitoring or input. With TermBridge, your Mac’s terminal session, whether it’s iTerm2, Terminal.app, or an integrated VS Code terminal, is mirrored securely to your mobile device. You can observe agent progress, review output, and intervene with commands as needed. This direct access keeps you connected to your development environment, ensuring you can guide your AI agent through its tasks without being physically present at your Mac.

Discussed on Personal blog (pushary.com)