View Markdown

Quickstart

Five minutes from here to your first notification and your first answered question, on your iPhone. You need an iPhone, and a computer with Python 3.9 or later as python3 (macOS and most Linux distributions have it).

1. Install the app

Install Kabu on your iPhone (Get the app, at the top of every page), open it and sign in. Allow notifications: they are how your agents reach you.

2. Install the kabu CLI

kabu is one Python file with no dependencies. Download it into ~/.local/bin:

mkdir -p ~/.local/bin && curl -fsSL https://usekabu.com/kabu -o ~/.local/bin/kabu && chmod +x ~/.local/bin/kabu

If your shell then says kabu: command not found, put ~/.local/bin on your PATH, here and in your shell’s profile:

export PATH="$HOME/.local/bin:$PATH"

3. Connect this computer

kabu login --device

kabu prints a code. In the Kabu app, open Connect a computer and enter the code it prints. This computer gets its own key, saved in ~/.config/kabu/config.json (readable only by you); you never copy or paste a key. The code lasts 10 minutes. Check the connection with kabu whoami.

4. Send an update

kabu send "Tests passed" "412 passed, 0 failed"

Your phone shows it. kabu prints one line of JSON, {"ok": true, "id": "msg_…", "status": "sent", "attachments": 0}, so scripts and agents can read what happened.

5. Ask a question

kabu ask "Deploy v1.4.2 to prod?"

kabu waits while your phone shows the question with Approve and Reject, on the Lock Screen too. Answer it, and the command ends with your answer: exit code 0 when you approve, 10 when you reject, 20 if nobody answers before it expires (5 minutes by default). Scripts branch on the exit code:

kabu ask "Deploy v1.4.2 to prod?" && ./deploy.sh

Offer options with -c, and read the pick from the JSON’s option:

kabu ask "Which environment?" -c staging -c production

Next

  • Let your coding agent do it. Claude Code and GitHub Copilot CLI: install the plugin, which teaches the agent when to send and when to ask. Other agents: give them MCP, or paste the output of kabu guide into their instructions.
  • Report a long run’s progress with sessions.
  • Call Kabu from your own service with the HTTP API.