AgentDock is a Mac utility that turns the permission prompts from several coding agents into one keyboard-driven Inbox: launched on Product Hunt on October 10, 2026, it keeps Claude Code, Codex and other terminal agents parked at the edge of your screen, shows which one is busy, and lets you answer every question and permission prompt from one place. It costs $19 once after a 7-day trial, and there is no account.
Key facts
| Fact | Detail |
|---|---|
| Question answered | AgentDock: One Inbox for Claude Code and Codex Approval Prompts |
| Short answer | AgentDock launched on Product Hunt on October 10, 2026. |
| Reading time | ~2 min |
| Last updated | 2026-10-11 |
What it ships with
- One Inbox. Questions and permission prompts from all of your agents land in a single list instead of scattering across terminal tabs.
- Answers typed into your terminal. Replies go straight into cmux, tmux, iTerm, WezTerm, kitty or herdr, so AgentDock is a layer over the terminals you already run rather than a new sandbox.
- At-a-glance status. You can see which agent is working without switching windows.
- Any-app messaging. The listing says the shortcut Option Option lets you message any agent from any app.
- Recaps. Finished work gets a short recap, written locally.
- Launch numbers. Day rank 33, 62 points and 28 followers at the time of writing, filed under Mac, Developer Tools and Artificial Intelligence.
Why the Inbox matters more than the terminal
The real cost of running several agents in parallel is not tokens, it is interruptions. Every agent that stops for permission pulls you back into its own window, and the usual workaround, approving everything up front, is exactly the habit that gets people into trouble. Collecting the questions in one queue keeps the human decision in the loop while removing the window-hopping tax.
Two caveats. First, it is Mac-only at launch, so this is not a Windows answer. Second, “nothing leaves your Mac” is the vendor’s own claim on the launch page, and it is worth understanding what the tool talks to before you point it at a work repository.
How this fits a Windows desktop agent
AgentDock is a control surface for agents you already run; it does not execute the work. If what you want is the other half, a Windows agent that proposes a plan, asks for your approval and then does the file, command and repetitive-task work on your own machine, that is what MOOGH does. The two patterns stack rather than compete, and the Windows-native side is a free download.
What to watch next
- Whether the Mac-only tag holds or a Windows build appears.
- Whether the Inbox stays useful past three or four agents, or becomes another queue to drain.
- How the one-time licence is maintained as the underlying agent CLIs keep changing.
FAQ
- Do I still need the terminals? Yes. AgentDock sits on top of cmux, tmux, iTerm, WezTerm, kitty and herdr rather than replacing them.
- Is my code sent anywhere? The launch page says there is no account and nothing leaves your Mac; check that against your own requirements before use.
- Can it run on Windows? Not at launch, the listing is Mac-only.
Related reading: