deck██████╗ ███████╗ ██████╗██╗ ██╗ ██╔══██╗██╔════╝██╔════╝██║ ██╔╝ ██║ ██║█████╗ ██║ █████╔╝ ██║ ██║██╔══╝ ██║ ██╔═██╗ ██████╔╝███████╗╚██████╗██║ ██╗ ╚═════╝ ╚══════╝ ╚═════╝╚═╝ ╚═╝
The Mac app for coding with AI agents.
Deck runs Claude Code, Codex, Gemini and other coding agents in one window, beside your terminals and a browser. You see every step an agent takes, approve what it wants to do, undo what went wrong, and test the result right next to it.
- agents
- claude code, codex, gemini cli, grok, opencode, cursor agent, amp, aider
- runs on
- macOS 13 or later, Apple silicon and Intel
- your code
- stays on your Mac. Nothing goes through our servers.
- read
src/settings/Page.tsx - edited
src/settings/Page.tsx+24 −3 - edited
src/theme.ts+11 - ran
npm test -- settings - opened
localhost:3000/settings
~/code/web ❯ npm test -- settings PASS src/settings/Page.test.tsx ✓ shows the dark mode switch ✓ keeps the choice Tests: 2 passed, 2 total
One window where coding agents do their work, and you stay in control.
AI coding agents like Claude Code and Codex are programs that read your code, run commands and edit files for you. Usually they run in a terminal, and their work scrolls past as text. Deck is the app you run them in instead. The window above is Deck; these are its parts:
Everything you have open
Each chat, terminal and page, with what every agent is doing right now, and a badge when one needs you. Tasks show here too.
A conversation with an agent
Say what you want. Each command and edit appears as it happens, and you approve, steer or undo it right there.
Your shell, beside the agent
Run the tests yourself, or run an agent's own interface. Fast, with search and splits.
Your app, beside the agent
Open your dev server next to the agent changing it, and see the result without switching apps.
For developers who hand work to AI agents.
- You use Claude Code, Codex or Gemini every day
and spend half the time scrolling a terminal to work out what the agent just did. Deck lists every step with its diff, as it happens.
- You keep several agents busy at once
and lose track of which tab is doing what. Deck gives each one its own branch and shows them all in one sidebar, with a badge when one needs you.
- You build things you can see
and check the page after every change an agent makes. Deck opens your app in a browser right beside the agent changing it.
- You don't use agents yet
Deck is still a fast terminal and browser in one window, and the easiest place to start: install Claude Code or Codex, open Deck, start a chat.
Any time you hand work to an agent.
| when you're | you | deck |
|---|---|---|
| building a feature | describe it in plain words | shows each file it reads and changes, runs your tests beside it, opens the page |
| fixing a bug | paste the error or a screenshot | hands it to the agent and shows the exact lines it changes |
| doing several jobs | start each one as a task | gives each agent its own branch, tracks them, lets you review and merge each |
| trying something risky | let it try the big refactor | saves your files first, so one right-click puts everything back |
| stepping away | go and get a coffee | notifies you when an agent asks for something or finishes |
| just working | use it as your terminal | is a fast terminal and browser on its own, and can be your default |
From install to a change you trust, in four steps.
- install
Install Deck and sign in. It finds the agent CLIs you already have, signed in with your own accounts.
- ask
Start a chat in your project with ⇧⌘N, pick an agent, and say what you want in plain words.
- watch
Every step shows up as it happens. Approve what it asks, send it more as it goes, or stop it with esc.
- keep
See every changed file, test it beside you, then keep it, merge it, or put it all back.
Why not just run the agent in a terminal?
| plain terminal | deck | |
|---|---|---|
| seeing what it did | scroll back through its output | every step listed, with its diff, as it happens |
| approving a command | answer a prompt somewhere in the scroll | a clear request above where you type |
| a turn that went wrong | fix it by hand, or reset with git | right-click your message, undo to here |
| several agents at once | a tab each, editing the same files | each on its own branch, all in one sidebar |
| checking the result | switch to a browser, find the tab | the page opens beside the agent |
| when it needs you | keep checking back | a notification, a badge and a Dock count |
| prompts you repeat | type them again | snippets, ;keywords and prompt rewriting |
Works with the agents you already use.
Deck has no AI of its own. It runs the agent CLIs you have installed, with your own accounts, so you keep the models and plans you're used to. Switch agents in the middle of a conversation and the new one picks it up.
What's inside.
;keywordman deck # every feature, and how to use it
One plan, with everything in it.
Deck Pro is every feature, on every Mac you sign in on, with updates installed for you. Price: . Your agents are billed by their own providers, as they are today.
Questions.
Is Deck an AI?
No. Deck is the app you run AI coding agents in. It uses the agents you already have, like Claude Code or Codex, signed in with your own accounts, and makes their work easy to follow and control.
Do I need Claude Code or another agent?
To get the most from Deck, yes: at least one coding agent CLI, such as Claude Code, Codex, Gemini CLI, Grok or OpenCode. Without one, Deck is still a fast terminal and browser.
Does it replace my code editor?
No. Keep your editor for writing code yourself. Deck is where agents do their work, and where you watch, approve, test and review it.
Is my code sent anywhere?
Not to us. Your agents run on your Mac and talk to their own providers directly, as they would in a terminal. Deck's server only knows your account and your plan.
Which Macs does it run on?
Any Mac with macOS 13 Ventura or later, Apple silicon or Intel.