PandaNpc lets you manage your own provider connections and model entries. Bring your own key (BYOK) means you supply the credentials required by the service you choose. A saved model entry describes a configuration; it does not purchase, host, or authorize that model for you.
Use this page to understand the list before following API Key and Base URL configuration.
Providers and model entries
A provider and a model are separate records:
| Record | What it describes | Example |
|---|---|---|
| Provider | Connection address, protocol and authentication settings | Your company API gateway |
| Model | Display name, upstream model ID, provider and optional limits | A model available through that gateway |
One provider can have several model entries. The display name helps you recognize an entry. Model Value must match the identifier accepted by the upstream API; changing the display name does not change the actual model.
What preset entries mean
The configuration screen can include provider or model entries marked Official. These are presets to help you configure a connection, not a promise of free access or included credits.
You can configure credentials for a preset provider, or copy a provider or model to create your own editable entry. Copying a provider leaves its credentials for you to supply. A model appearing in the list does not prove that your account has permission to call it.
Manage your model list
Open Model Configuration and select a provider to see its models. Add a model with its display name, Model Value, type and provider. Custom entries can be edited, copied, enabled or disabled, and deleted through their available actions.
Keep labels distinct when you use several accounts or endpoints for the same model. For example, a work gateway and a personal provider may expose the same model ID while using different credentials and limits. Associate each entry with the correct provider.
A default model is a configuration preference for the features that use it. It does not automatically change every existing remote Agent session.
Context window and output length
Context Window records the model's supported context capacity. Max Tokens records an output limit. Use limits documented for the exact API and model you selected.
An empty field means you have not supplied that value. It does not mean PandaNpc has verified the upstream capacity. Defaults and the way limits are applied depend on the consuming feature or engine. A larger number cannot increase what the upstream model accepts.
Similarly, choosing a model type does not add image input, tool calling or other capabilities to the model.
Verify availability in the feature using the model
After saving an entry, select it in the feature that uses this configuration and send a short text request. Check that the expected provider handled it. Test images, tool calls and long context separately if you need them.
A successful save proves the configuration was stored. A successful text reply proves that particular request worked. Neither proves compatibility with every PandaNpc feature or every upstream capability.
Authentication, model availability, quotas and service limits belong to the configured upstream service. Check its current settings instead of relying on a fixed model catalog in documentation.
Remote Agent models are configured separately
Claude Code and Codex run on the connected computer and use their engine or session configuration. Updating this list does not automatically replace the model of a running process. For engines or tasks that accept a PandaNpc model entry, select it through that engine's configuration and verify the resulting session.
See the PandaPaw host guide for remote execution and accounts and model costs for the separation between PandaNpc membership and model-provider usage.