Connect your own models
Add your own API key or an in-house inference server (vLLM, Ollama, and others) to the Keural model list.
In the desktop app, go to Settings → Models → Gateways and connect an OpenAI-compatible or Anthropic-compatible server. Then you can select that server's models in the Gateway group of the level menu in the input box.
How to connect
- Click Connect gateway.
- Choose a preset (vLLM, Ollama, OpenAI, Anthropic, and so on) and enter a name. If you choose manual entry, it connects as OpenAI-compatible.
- Enter the Base URL as your model provider describes. Do not add the API endpoint path.
- Enter the API key. Leave it blank for local or internal network servers that do not require authentication.
- Click Check connection to load the model list. Choose the model you want and click Add.
Where it is saved
The save location is determined by the address. A public https address is saved to your account and can be used on other devices signed in to the same account. A localhost or internal network address is saved only on this computer and shows the Saved on this device only label.
The API key is not shown again after you save it.
Billing
Usage fees for models connected with your own API key are charged by that provider and are not included in Keural usage.
Models on a connected server follow that server's policies. To use this within an organization, first confirm that the server meets your internal security policy.