Run AI inference on your own Gcore account

What connecting it does for you

Connecting Gcore Everywhere Inference brings your inference deployments — the model endpoints you run on your own Gcore account, and what each one is serving — and the GPU options available to you into your Zinn® dashboard.

Before you start

A Gcore account of your own, with Everywhere Inference enabled on it. This is bring-your-own only: we do not resell Everywhere Inference, so the product, the contract and the bill are between you and Gcore, and what you run there is charged to your Gcore account, not to your Zinn® plan.

1. Create the key at Gcore

In the Gcore Customer Portal, open your avatar at the top right, then Profile → API tokens → Create token. Name it, choose Never expire or an expiry date, and give it a role for each product it needs. Gcore shows the token once; copy it then.

The same token works for every Gcore product you connect, so if you have already connected one Gcore product you can paste the same token again — provided its roles cover this product too. A token cannot be given a higher role than your own login has.

Give the token a role for Cloud. You also need the numeric ID of the Gcore Cloud project your deployments live in — Gcore's inference service is organised by project, so a token alone cannot list them. It is shown for the project in the Gcore Cloud section of the Customer Portal.

2. Connect it here

Open Integrations in your dashboard and choose Connect an account. Pick AI inference as the group and Gcore Everywhere Inference as the account, fill in API token and Project ID, and press Connect account.

We test what you paste before anything is saved. A key that does not work is never stored, and the answer says what was wrong with it. A key that works is kept encrypted in our secrets vault — never in our database — and is never shown again, not even to you.

What happens next

• Your deployments and the GPU options your project offers are listed under **Your

own services, reached from the connection's Open your own services button on Integrations**.

• Deployments run and are billed on your Gcore account, under your contract with Gcore.

If it does not connect

It says the product is not enabled. The token is valid, but Everywhere Inference is not switched on for that Gcore account — Gcore then refuses every read. Enable it in the Gcore Customer Portal (or ask Gcore to), then connect again. A token whose role does not include Everywhere Inference produces the same refusal; create one that does.

It connects but lists nothing. The Project ID points at a project with no deployments, or at a project the token cannot reach. Check it is the project your deployments are in.

It says the key was rejected. Almost always one of three things: a space or a line break copied with it, a key that has expired, or a key that was revoked or regenerated after you copied it. Create a fresh one and paste it again.

It connects, but something later fails. The key authenticates but lacks a permission the action needs. Create a new key with the permissions listed above, then disconnect the old connection and connect the new key.

Disconnecting

Open Integrations, find the account and press Disconnect. That deletes the stored key at once. Anything that was using it stops at its next action, and the screens that depended on it say so rather than failing quietly.

Disconnecting does not undo what was already done — records, deployments or settings we changed on your account stay as they are. If you think the key itself may have leaked, also revoke it at the vendor; disconnecting removes our copy, not theirs.