A terminal for Claude Code

coterm is a native macOS terminal built for running AI coding agents, and Claude Code is a first-class fit. coterm is just a terminal, so claude runs in any workspace out of the box, and the things that make running agents painful, keeping track of many at once and noticing when they need you, are what coterm is built for.

Run many Claude Code sessions, organized

Open a workspace per task and run Claude Code in each. The vertical sidebar shows every workspace with its git branch, directory, ports, and the latest line of Claude's output, so a dozen parallel sessions stay legible instead of buried in tabs.

Notification rings when Claude needs you

When Claude Code finishes or asks for input, the pane rings and the sidebar shows an unread badge, so you can let several agents run and come back to the one that needs a decision. Notifications fire automatically, and you can also trigger them from Claude Code hooks.

Claude Code teams as native panes

coterm runs Claude Code's teammate mode with one command, and the teammates spawn as native coterm splits with their own sidebar metadata and notifications, no tmux required. See the Claude Code teams docs.

Check on Claude from your phone

coterm has an iOS companion app (beta): pair your iPhone with your Mac and check on your Claude Code sessions, with optional notification forwarding, while you are away from the desk.

Scriptable

Every action is available through the coterm CLI and a Unix socket: create a workspace, launch Claude Code in it, send input, read the screen, and drive an in-app browser to verify changes, all from a script.