AgentDeck CLI
A local dashboard for Claude Code and Codex sessions: who is working, who waits for you, plus a 3D aquarium and Stream Deck
- 286 ★ on the repository
- Last commit
- MIT
Extended access
Why this level
- Writes hooks into Claude Code and Codex settings and keeps a background daemon that sees the events of every tool call
- Reads session logs and transcripts, and the panel can answer permission requests, that is, approve agent actions
Install
- Works in
Manual install
npx @agentdeck/setupRequires Node.js 22, 24 or 26. Then run agentdeck daemon install.
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- Writes hooks into Claude Code and Codex settings and keeps a background daemon that sees the events of every tool call
- Reads session logs and transcripts, and the panel can answer permission requests, that is, approve agent actions
This is third-party code. Review the repository files before installing.
What it does
AgentDeck gathers all running AI agents into one panel and shows which is working, which waits for a decision and which has finished. It runs on Mac, in a terminal, on Windows 11 and on Linux. For the terminal it installs a daemon that connects Claude Code and Codex CLI lifecycle hooks and reads Codex logs. For opencode an observer plugin is used, for Kiro transcript reading, and Hermes Agent connects as an opt-in read-only preview. Install with `npx @agentdeck/setup`, then `agentdeck daemon install` refreshes the hooks and starts observation. `agentdeck dashboard` opens the terminal panel. It shows session names, current tools, a shared timeline and links between subagents and tasks. Permission and input requests stand out, and you can answer them on surfaces where the session has a real control path. Elsewhere prompts are display-only. In the 3D scene each session becomes a creature whose movement follows the agent's activity. Usage gauges for Claude, Codex and z.ai subscriptions appear when the provider supplies usage data. You can add an iPad, iPhone, Android, Stream Deck, Ulanzi, e-ink and ESP32 displays to the Mac.
Who it is for. People who keep several agent sessions open across projects and want to see at once which one is waiting for an answer.
Good fit when
- Five Claude Code and Codex sessions run at once and you must notice which one holds a permission request
- You want to see the remaining Claude or Codex subscription limit next to your work without opening settings
- A Stream Deck or tablet sits nearby and you want quick-answer buttons for the agent
- A task runs for an hour and you step away: status and completion show on a separate screen
- You need a shared timeline of what each subagent launched and how it ended
Not a fit when
- You work with one session in one window: a separate panel adds little
- You only need token and cost accounting: `ccusage` or AI Observer is simpler
- Your team forbids hooks in agent settings or reading session logs
- You run macOS below 15 or a Node.js version outside 22, 24, 26 for terminal mode
Example request
Install AgentDeck, connect observation for Claude Code and show a terminal dashboard with all my current sessionsLimitations
- Terminal mode needs Node.js 22, 24 or 26 and macOS 15 or newer, Windows 11 or Linux. The Mac app needs macOS 26 and iPhone and iPad need iOS 17.
- You can answer the agent only where the session gives a real control path. Otherwise requests are shown without an option to answer.
- For Kiro and some other agents observation works from transcripts, it lags and shows idle only.
- Release 2.0 ships channel by channel: the macOS version was still awaiting review when the README was written, and the Stream Deck and Ulanzi marketplaces are tracked separately.
How to disable. Stop and remove the daemon with the commands from `agentdeck` help (the daemon section), and remove the hooks from Claude Code and Codex settings if installation added them. `agentdeck diag agents` shows connection diagnostics. The Mac, iPhone, iPad and Android apps are removed the usual way, as are the Stream Deck and Ulanzi plugins.
FAQ
How do I install AgentDeck for the terminal?
Run `npx @agentdeck/setup`, then `agentdeck daemon install`. Start your agent as usual, for example `claude`, `codex`, `opencode` or `kiro-cli`, and open the panel with `agentdeck dashboard`.
Which agents are supported?
Claude Code connects through lifecycle hooks and is the primary integration. Codex CLI works through hooks and logs, and Codex Desktop is observed on macOS. For opencode there is an observer plugin, for Kiro transcript reading, and Hermes Agent connects as a read-only preview.
Do I need hardware beyond a computer?
No. A Mac or a terminal works on its own, and a tablet, Stream Deck, e-ink and ESP32 displays are optional additions.
Can I answer agents from the panel?
Where the session has a real control path, yes. Without one, permission and input requests are display-only.
What if an agent does not show in the panel?
Run `agentdeck diag agents`; it shows the connection state of each agent. Details are in the installation and troubleshooting documents in the repository. For Codex, confirm hook trust when prompted.
Discussion
No questions or reviews yet.
Related
CLI toolsA self-hosted control center for coding agents: Claude Code, Codex, Copilot, OpenCode and Pi from your phone, desktop and terminal
Native macOS terminal with an MCP server and CLI: agents see panes, commands and output, and you grant permissions in one window
Agent of Empires
Agent of Empires (AoE)
A session manager for coding agents: runs Claude Code, OpenCode, Codex and other CLIs in parallel, isolated tmux sessions with a TUI and web dashboard
AgentBox runs Claude Code, Codex and Open Code in separate virtual machines on your PC, a self-hosted server or a cloud.
A Docker sandbox for coding agents: a Debian image with Copilot CLI, Codex and Pi, and SSH, RDP and Docker enabled by flags.
Hookdeck CLI forwards webhooks to a local server, manages Event Gateway and runs an MCP server through which an agent investigates delivery failures
Comparison with similar tools
Cards that share the main search query, and how they differ.
| Tool | Official | Repository stars | Main difference |
|---|---|---|---|
| AgentDeck (this card) | no | 286 | |
| Agent Deck | no | 1,216 | a terminal session manager with a status list, while AgentDeck adds a 3D scene, subscription gauges and external displays |
| Claude Code Agent Monitor | no | 1,152 | a web dashboard for Claude Code and Codex, while AgentDeck adds hardware controls and mobile apps |
README in short
The README opens with links to app stores, npm and the website, then describes the state of the 2.0 release by channel. It continues with interface choice, installation for the terminal, Mac and mobile devices, a feature list, a table of supported agents, building from source and device sections.