Skip to main content

Your first endpoint

You need an Engress account, one endpoint, and a connect token before the agent can register.

1. Sign in and create an endpoint​

  1. Open engress.io and sign in.
  2. Create a new endpoint. Note the endpoint ID (a UUID shown in the dashboard).
  3. Issue a connect token for that endpoint. This plaintext token is shown once. Copy it before closing.

Store the token somewhere safe — you cannot retrieve the plaintext value again. If you lose it, revoke the old token and issue a new one.

2. Start your model server​

Run Ollama, LM Studio, or any HTTP server on a local port. The examples below use Ollama on port 11434:

ollama serve

3. Run the agent​

Point the agent at your endpoint and local backend:

engress-agent tunnel serve \
  --edge-host tunnel.engress.io \
  --edge-port 7443 \
  --endpoint-id "<your-endpoint-id>" \
  --local 127.0.0.1:11434

The agent dials outbound to the edge on port 7443. It does not open a listener on your public IP.

When registration succeeds, the dashboard shows your endpoint as connected and displays the public hostname (for example quiet-otter.edge.engress.io).

4. Call the public URL​

curl "https://quiet-otter.edge.engress.io/v1/models"

Replace the hostname with the one assigned to your endpoint. Traffic flows: client → Engress edge → outbound tunnel → your local port.

Token reminder​

This plaintext token is shown once. Copy it before closing.

Next​

Connect tokens — scopes, rotation, and revocation.