AgentPet CLI
A desktop pet and menu-bar monitor that shows which coding agent is working, done or waiting for your input
- 399 ★ on the repository
- Last commit
- MIT
Broad access
Why this level
- The Install button writes hooks into the settings of Claude Code, Codex and other agents, and without Remove they stay after the app is gone
- For subscription limits the app reads your Claude and Codex sign-ins
- After GitHub sign-in token statistics and the last folder name of each project go to the website
Install
- Works in
Manual install
brew install --cask ntd4996/tap/agentpetThe macOS option. For Windows and Linux the installers are on the project Install page and in the releases. After installing, click Install next to your agent in Settings.
Access and checks
How we reviewBroad accessThe tool can change data, run commands or handle secrets.
- The Install button writes hooks into the settings of Claude Code, Codex and other agents, and without Remove they stay after the app is gone
- For subscription limits the app reads your Claude and Codex sign-ins
- After GitHub sign-in token statistics and the last folder name of each project go to the website
This is third-party code. Review the repository files before installing.
What it does
AgentPet is a desktop app for macOS, Windows and Linux with an `agentpet` command that shows the state of coding agents. A menu-bar monitor lists agents with their project, current action and a timer. The icon turns orange with a count when an agent waits for your input, and a pixel pet on the desktop reacts to events. The macOS app is written in Swift and SwiftUI, and Windows and Linux share one Tauri app. Claude Code, Codex, Gemini CLI, opencode and Factory Droid report the states working, waiting for input and done. Cursor, Windsurf, Antigravity, GitHub Copilot, Kiro CLI and Pi give only working and done. GLM goes through Claude Code. The Install button in Settings puts hooks into agents, and `agentpet run -- <command>` wraps any other command: the agent counts as working while it runs and done after it exits. A daemon on a Unix socket receives the events. The pet feeds on the tokens agents burn and on finished sessions, subagents included. It earns experience, grows through five stages and unlocks 14 achievements. The app is fully local: the web profile, leaderboard and cross-device sync switch on when you sign in with GitHub. Session history is kept locally for 90 days.
Who it is for. Developers who run several coding agents at once and want to see which one is working, which is done and which waits for an answer.
Good fit when
- Three agents work in different repositories and you need to know which one waits for your permission
- You want a notification when an agent finishes a long task or stops with a question
- You want to see which projects and agents burn tokens: the web dashboard counts by day and month
- An agent without a ready hook runs through `agentpet run` and the monitor still sees it
- Each project needs its own pet and experience should accumulate per project folder
Not a fit when
- You need a token report without signing in on the website: the dashboard opens after sign-in and connecting
- A GNOME desktop without AppIndicator: the tray icon will not show until the extension is installed
- Strict data requirements: after sign-in, statistics and the last folder name of each project go to the website
Example request
Start a long task in Claude Code and let me know when it finishes or needs an answerLimitations
macOS 13 or newer, Windows 10 and 11 or Linux are required. The Windows and Linux builds are unsigned, so SmartScreen warns on first launch. On macOS the app is signed and notarized.
Cursor, Windsurf, Antigravity, Copilot, Kiro CLI and Pi agents show only working and done.
The app reads Claude and Codex subscription limits straight from your sign-ins. The web profile, leaderboard and cross-device restore require signing in with GitHub.
How to disable. In Settings, General, click Remove next to each connected agent: this strips the AgentPet hooks. Then remove the app and its data: `brew uninstall --cask agentpet` and `rm -rf ~/.agentpet`.
FAQ
How do I install AgentPet?
On macOS with `brew install --cask ntd4996/tap/agentpet` or from the AgentPet.dmg file in the releases. For Windows download the installer or the portable build, and for Linux use the AppImage, `.deb` or `.rpm`. Then click Install next to your agent in Settings.
Which agents are supported?
Eleven: Claude Code, Codex, Gemini CLI, opencode, Factory Droid, Cursor, Windsurf, Antigravity, GitHub Copilot, Kiro CLI and Pi. GLM works through Claude Code, and any other command runs through `agentpet run`.
What does the app send over the network?
Nothing until you sign in with GitHub: the app runs fully locally. After sign-in the web profile, leaderboard and cross-device restore appear. Only the last folder name of a project goes to the usage dashboard, never the full path.
Where does the pet get experience?
From the tokens agents burn and from finished sessions, subagents included. Each pet keeps its own level, and a project can be bound to its own companion.
Where do new pets come from?
From the online gallery: you add a pet in Settings, Pet, or on the website. The app bundles no pet art, and every pack loads at runtime and belongs to its author under their own license.
Discussion
No questions or reviews yet.
Related
CLI toolsA desktop pet that reacts to Claude Code, OpenCode and Cursor through hooks and MCP without putting code, paths or secrets in its speech
A gallery of animated companions for coding agents: a CLI installs a pet in one command, and a desktop app shows it reacting to the agent's activity
The /myoffice command opens a browser office showing what every agent on your machine is doing: thinking, waiting for you or done.
Agent Standup shows Claude Code, Codex and Mistral Vibe sessions as a live team call: who works, who waits.
Mac screen HUD for agents: quotas, tokens and sessions of Claude Code, Codex and more, plus a built-in MCP server.
Desktop pet with a CLI and a skill: an agent keeps sticky notes on the desktop while the console calls the McDonald's MCP service
Comparison with similar tools
Cards that share the main search query, and how they differ.
| Tool | Official | Repository stars | Main difference |
|---|---|---|---|
| AgentPet (this card) | no | 399 | |
| OpenPets | yes | 1,388 | a pet on hooks and MCP for three agents with no usage dashboard and no growth from tokens |
| Claude Code Agent Monitor | no | 1,152 | a web dashboard of Claude Code and Codex sessions, with no pet and no menu-bar notifications |
README in short
The README opens with a headline and links to the website, gallery and leaderboard. A summary of the monitoring and pet features follows, then a table of supported agents and screenshots. Installation is covered per system, and after it come first steps, uninstalling, the usage dashboard, the pet format, platforms, community ports and a contributor list.