The setup wizard

The wizard runs once, right after install, and does three things: reads your hardware, recommends a backend and model that'll actually run well on your machine, and gets that backend running.

Hardware detection

It checks your CPU, RAM, and GPU (including available VRAM) and uses that to recommend a model tier — there's no guesswork or manual model-picking required. If your machine doesn't have much GPU memory, it'll steer you toward a smaller, faster model instead of one that'll crawl or fail to load.

Local model vs. Claude API

You get a choice of backend:

• Local (Ollama) — runs entirely on your machine, no API key, no per-token cost, and nothing you type ever leaves your computer.
• Claude API — you bring your own Anthropic API key; noticeably more capable, at the cost of a usage-based bill from Anthropic (not from us).

You can switch backends later from the dashboard's Settings panel — this isn't a one-time decision.

Getting Ollama running

If you chose local and Ollama isn't already running, the agent starts it for you automatically — you shouldn't need to do this by hand. If that fails (usually because Ollama isn't installed yet, or is installed somewhere the agent can't find it), see Ollama won't start.

Next step

Once your backend is running, the wizard asks for your license key — see Activating your license.