Sign up, drop the agent on a GPU box, hit the gateway. That's the whole tutorial.
Go to app.huxkan.com/signup. Free 14-day trial, no credit card.
From the console: Gateways → New fleet. Give it a name (e.g. prod-us-east). Kan generates a fleet enroll token and shows you the installer command.
SSH into the GPU VM and paste the installer:
curl -sSL https://install.huxkan.com | FLEET_TOKEN=FLEET_ETK_live_… sh
The installer puts the agent at /usr/local/bin/kan-agent, registers a systemd unit, starts it. The agent enrolls with Kan, brings up APISIX/Redis/Prometheus/Grafana/Loki containers, and starts heartbeating.
From the console: Gateways → your fleet → globe icon. Type the address(es) your end users will reach the gateway at: 10.0.1.50, gateway.acme.com, or both comma-separated. The agent uses this for the SNI list and cert SAN.
If you already have vLLM running on the box: Gateways → your node → Discover & Adopt. Hux scans for OpenAI-compatible servers and registers them.
Or, from Models → Deploy new, pick an HF repo, pick a node, click Deploy. Hux pulls and starts the container.
Keys → Generate. Pick a consumer name, optionally assign a consumer group with rate limits / budgets. Copy the sk_… key (shown once).
curl https://gateway.acme.com/v1/chat/completions \
-H "Authorization: Bearer sk_…" \
-H "Content-Type: application/json" \
-d '{"model":"llama-3.2-1b","messages":[{"role":"user","content":"hello"}]}'
You should see a chat completion stream back. Watch the Dashboards page in Kan — the request shows up in real-time metrics.