Getting Started

OwnLLM has two halves:

• A SaaS control plane at https://ownllm.app that handles auth, billing, audit, and the public OpenAI-compatible API.
• A local agent that runs on your GPU machine and serves inference. It ships either as the Atlas desktop app (Mac, Windows, Linux GUI) or as the ownllm CLI (headless).

Pairing the local agent to your tenant takes one command. After that, everything — model selection, audit logs, API keys — happens from https://ownllm.app.

The five-minute path

1. Sign up at ownllm.app and pick a plan.
2. Copy your pairing key from the onboarding wizard (ownllm_pair_<32-hex>, valid for seven days, single-use).
3. Install the agent on your GPU host:
   • Desktop: download the .dmg / .msi / .deb from /download.
   • CLI: curl -fsSL https://ownllm.app/install.sh | sh.
4. Pair the machine:
   • Desktop: paste the key into the first-run wizard.
   • CLI: ownllm pair ownllm_pair_<key>.
5. Open https://<your-slug>.ownllm.app in any browser, sign in, chat.

Once paired, the agent opens an outbound Cloudflare Tunnel and registers itself in your tenant. No port forwarding. No DNS. No reverse proxy.

What runs where

Your prompts and completions transit ownllm.app (so we can authorize, log, rate-limit) but inference runs on your hardware. We never send model weights or completions to a third-party provider.

Next steps

• The CLI reference (lands in the next release) covers pair, service, models, runtime, recipes, and doctor.
• The Desktop app guide (next release) walks through the pairing UI, dashboard, and tray icon.
• The API docs (next release) explain OPENAI_BASE_URL, key scopes, budgets, and supported endpoints.