Source: https://harnsy.dev/features/terminals

> Claude Code, Codex, OpenCode and pi agents in tabs of tmux or harnsy’s own terminal host: watch and type into any terminal from the dashboard and reopen the team after a reboot.

# Watch any agent. Step in when needed.

Agents run in tmux if it is installed, otherwise in harnsy’s own terminal host (on Windows, always its own), a tab each. The dashboard opens their terminals, reopens the team after a reboot and lists every agent with its role and context.

Every agent in its own tab; the amber ones wait for you. Sample data.

### A tab per agent.

Agents run in a tab of their own (tmux, or harnsy’s own terminal host), named after the agent, so you find the one you need at a glance.

### The lead opens agents.

The lead starts a new agent with its first task already in it, from a ready profile: harness, model and flags.

### Back after a reboot.

Closed the terminal or restarted the machine? “Reopen disconnected” on the team returns every agent to its own session: its history and its role.

### The live terminal.

Open any agent’s terminal over the page, full screen if you like, and watch it work. Switch to typing to step in; Ctrl+Shift+L flips between watching and typing.

### Tabs that wait for you.

A tab whose agent waits for a decision gets an amber underline, and Ctrl+Shift+F jumps to the next one. A coloured dot shows the harness.

### In your own terminal.

One copied command resumes any agent’s Claude Code, Codex or OpenCode session in your terminal. If it still runs, harnsy says where to close it first: no two processes on one session.

### Every agent on one screen.

Who works, who waits, in which role and harness, and how much context is left: Claude Code, Codex and pi report it, OpenCode shows none. A profile opens in a side panel.

### Finish an agent in one step.

One action closes its tabs, releases its seats, notes its open tasks and tells the lead. Tabs of an agent that left pass to its successor, or to the lead.

### Put an agent to sleep, or turn it off.

Sleep closes the agent’s tab but keeps its seat and its session, so it can come back where it stopped. Turn off frees the seat for good, and you do it in the dashboard. Both work for one agent or for a whole team from the dashboard. In `harnsy term`, with tmux or harnsy’s own terminal host, you put one agent to sleep and wake it with the keys s and w.

### See what an agent is running.

When an agent builds or tests with `harnsy run`, its row shows the command, how far it has got (read from its output) and how long it usually takes (from its past runs). A heavy run started with `harnsy run --heavy` waits until the machine has room, so several agents do not stall it together.

### A console only you open.

The Console is a plain terminal in the project’s folder, next to the agents. Only you can open it from the dashboard; no agent does.

Rebooted, and spent half a day remembering which agent was doing what where.

## One button, &quot;Reopen disconnected&quot;, brings the whole team back.

Howharnsy opens a terminal per role, starts the harness and gives each agent back its own session.

Reopen disconnected

## harnsy term

All the agents you work with across your teams, in one place, in your terminal.

### Who waits for you is on top.

The list groups agents by project and shows each one’s role and status. An agent that waits for your answer goes first, with an amber dot, so you see it without hunting through tabs.

The console: agents by project on the left, the chosen agent’s live screen on the right. Sample data.

### The live screen, with typing.

The right side is the chosen agent’s own terminal, live. Type, and it goes to that agent. To only look, press Ctrl+\], then r: while you only look, nothing you press reaches the agent. Press Ctrl+\], then r again to type.

### Click, or Ctrl+\] and a key.

Click an agent to switch to it, drag over the screen to copy text, and use the wheel on the list or on the history. For keys, press Ctrl+\] first, then the key you want; the table has them all.

Mouse and keys

### Close it, and the agents keep running.

The agents run in harnsy’s own terminal host. Closing the console, or restarting harnsy, does not stop them. To leave the console, press Ctrl+\], then q; the agents stay where they are.

### Opening and closing agents stays on the dashboard.

harnsy term is for watching and answering. To open an agent or close one, use the dashboard, as before.

### Part of harnsy.

harnsy term comes with harnsy. It needs no key.

### Everything opens as a window.

In `harnsy term` agents, usage, a task and an ordinary shell each open as a window, and the layout is kept between runs. A window “Waiting for you” holds the same questions as the dashboard: answer with a click or a key; an agent with a question is marked ⧗.

## Limits, honestly

- Opening agents, the live terminal and “Reopen disconnected” need a terminal harnsy can drive: tmux if it is installed, otherwise harnsy’s own terminal host (always its own on Windows, 10 (1809) and later). Without a terminal, messages, teams and the dashboard still work. On Windows there are no splits and no separate windows: one tab per agent.
- Agents you start yourself, outside harnsy’s terminal, still message and work in teams; harnsy just can’t show or reopen their terminals.

[Several Claude Code sessions on one screen →](https://harnsy.dev/claude-code-multiple-sessions-dashboard)

## Your first team is one prompt away.

Installing takes one sentence to your agent. To run your agents, harnsy drives a terminal: tmux if it is installed, otherwise its own terminal host (on Windows, always its own).

`Install harnsy following https://harnsy.dev/llms.txt`

## Questions

What do I need first?

At least one harness installed and signed in. Nothing else to install: to open agents, harnsy uses tmux if you have it, otherwise its own terminal host (always its own on Windows).

Which systems?

Linux, Windows and macOS. On Windows harnsy opens agents in its own terminal host (conpty) on Windows 10 (1809) and later; the dashboard and the command line work as everywhere.

Do I need a special terminal?

No. harnsy picks one for your system: tmux if it is installed (Linux and macOS), otherwise its own terminal host, which is always the one on Windows. Without a terminal, the agents you open yourself still message each other and work in teams.

All questions →
