CLI

Terminal ops for reinforcement learning agents—train, trade, and automate from your shell. Not a coding agent. Not a second dashboard.

Train and trade from the terminal

The Kabu CLI is a focused ops surface for RL agent lifecycle: start and monitor training runs, deploy trading agents, and inspect shared performance metrics.

Train — start, stop, and monitor runs; inspect learning metrics
Trade — deploy and stop agents; positions, fills, and PnL
Automate — subcommands for scripts and CI; optional --prompt for one-shot agent turns
Insight — performance views from the same stats APIs as the dashboard (CLI only presents)

Deep experiment authoring, billing, org admin, and rich charts stay on the web dashboard. The CLI talks to the same API—it does not reimplement the product in a terminal.

Three ways to run

Pick the mode that fits: interactive shell, scriptable subcommands, or a one-shot prompt.

kabu — fullscreen app shell (header, panels, footer intents)
kabu <noun> <verb> — non-interactive commands for piping and CI (e.g. kabu runs list)
kabu --prompt "…" — one-shot agent turn for automation

Install

Published as @zimablue-io/kabu on GitHub Packages. Binary name: kabu. Requires Node 22+.

1. Auth to GitHub Packages (once per machine)

Add to ~/.npmrc. The PAT needs read:packages (and repo if the package is private to the org).

@zimablue-io:registry=https://npm.pkg.github.com
//npm.pkg.github.com/:_authToken=YOUR_GITHUB_PAT

2. Install globally (pick one)

npm:

npm install -g @zimablue-io/kabu

pnpm:

pnpm add -g @zimablue-io/kabu

3. Verify and sign in

kabu --help
kabu login

Login uses browser OAuth. The default API origin is https://www.usekabu.app. For automation you can also use an org API key or paste a bearer token.

Monorepo development

Contributors working in the Kabu monorepo can run the CLI without publishing.

just cli
# or
pnpm --filter @kabu/cli dev

Package name in the monorepo is @kabu/cli; the published name on GitHub Packages is @zimablue-io/kabu.

Next steps

Create an account, open the dashboard, or dig into training and live trading.

Get started · Sign in · Training · Live trading · Features overview