Working with Code Agents

Otty is built for the way people code now — with an agent doing the typing. Run Claude Code, Codex, or OpenCode as first-class panes and let Otty handle the rest: it shows you what each agent is doing, pings you when one needs you, and gives you real input and editing tools instead of a bare prompt.

If you live in a coding agent, this is the terminal for you.

Why Otty for agents

See every agent at a glance. Otty's vertical tabs carry live badges, so a backgrounded agent tells you whether it's still working, done, or waiting on you — without switching to it. → Monitor Tasks

Get pinged, not stuck watching. System notifications fire when a task completes or an agent is waiting for your approval or input — and Otty can keep the Mac awake so a long run doesn't stall on sleep. → Monitor Tasks

A real input surface. The Composer gives you comfortable multi-line editing, undo/redo, image paste, and a float-on-top panel — write a proper prompt instead of fighting a single line. → Composer

Line up the next step. Queue commands while an agent is busy (⌘⇧M) and Otty dispatches each one at the next idle prompt — stack a plan and walk away.

Pull context straight in. Send to Chat drops a terminal selection, the last command's output, or a file-pane snippet into the conversation — no copy-paste shuffle.

Pick any thread back up. Every conversation is captured. Resume a session days later or search across all of them — your context outlives the tab.

Explore in parallel. Fork or branch a conversation to try a second approach in a split or a new tab, without losing the original thread.

Check the work without leaving. Otty's built-in file viewer and editor lets you open and read what the agent changed right beside the chat — no separate editor window.

Quick start

Otty works with the coding agents you already run — it doesn't replace your CLI, it plugs into it.

1. Run your agent in an Otty terminal the way you always do — claude, codex, or opencode.
2. Otty offers to install a small hook (Claude Code, Codex) or plugin (OpenCode) into that agent's own config. Approve it once — nothing else about your setup changes; the hook just lets the agent tell Otty when it starts working, finishes, or needs you.
3. Restart the agent in that tab if prompted. From then on, tab badges, notifications, history, and fork all light up automatically.

See Setup for the per-agent details and how to manage or remove the integration.

See also

• Setup — providers, API keys, models.
• Composer · Prompt Queue · Send to Chat — getting input in.
• Monitor Tasks · History · Fork / Branch Session — staying on top of running and past work.