Get started with Kaidera OS

Before you start

Prepare a machine with Docker, Python 3.12 or newer, enough disk for local services and project data, and at least one model provider API key. Kaidera OS is designed for macOS and Linux first.

• Minimum practical machine: 2 CPU, 4 GB RAM, and about 50 GB of available disk.
• Comfortable local machine: 2 or more CPU, 8 GB RAM, and room for logs, caches, and project artifacts.
• Docker runs the Cortex stack and app database.
• Provider keys are added after install in Settings.
• A license key is required at runtime to unlock the features assigned to your account or edition.

Install

Use one public install path. Homebrew is the simplest path on macOS and Linux. npm and curl install the same current engenos CLI.

• Homebrew: brew install engen-ai/engenos/engenos, then run engenos install and engenos start. The package coordinate retains its legacy machine name during the compatibility window.
• npm: npm i -g @engenai/engenos, then run engenos install.
• curl: use the installer published from the current homebrew-engenos repository.
• All public install paths should converge on the same product and CLI.

Start the console

After installation, start the local console with the current engenos CLI. The console runs locally and opens in your browser. The standard local address is http://localhost:8765.

Confirm health

Before configuring workers, confirm the console, Cortex API, and app database are reachable. A healthy runtime prevents confusing empty graphs, missing history, or failed worker dispatch.

• Console: http://localhost:8765 should load the Kaidera OS UI.
• Cortex API: the memory backend should report healthy in Settings or the Dashboard.
• App database: Settings should show the store connected.
• Provider path: at least one configured model provider should be available before model-backed workers run.

First-run setup

A fresh install starts empty. The first-run flow collects the project name and key, workspace root, project scope, first lead worker, initial team template, cadence settings, and provider preferences.

What to do next

Bring one project online first. Keep propose mode on while you validate the project root, worker roster, provider settings, and Cortex memory. Turn on broader autonomy only after the first approved runs behave as expected.