AI providers and models

Open Connectors → AI to see the models and access modes available to your workspace. Availability depends on account access, the enabled catalog, and a usable credential path.

​

Choose an access mode

An owner or admin manages the workspace’s available access settings. Not every publisher supports a direct customer API key. Expanded model choices can be available through managed access without a separate native connector for each publisher.

​

Connect a provider

Choose a provider or publisher from AI and inspect its availability. Where Your API key is supported, open the provider’s credential settings, add the key, and check readiness. Where Erstan Managed is available, select that mode and verify the desired model is enabled. A key is a workspace credential, not a chat attachment. Replace or remove it from the credential settings when needed. The interface explains unavailable access modes or missing setup; changing a model does not automatically authorize a new funding mode.

​

Choose which models are enabled

The model catalog is curated by Erstan. Use the available workspace model controls to choose enabled models and a default. A model also needs an available provider and credential path before it can run. Changing a default does not rewrite the models saved in existing published Agents. Review and republish an Agent when its own model selection needs to change.

​

Workspace defaults and Auto

Open Settings → AI to configure the default chat model or Auto, Auto preferences and limits, and a separate Logic default for new AI Branch and AI While steps. Auto chooses from eligible workspace models according to the configured preferences and limits. It does not grant access to disabled models. A concrete model remains useful when you need a particular capability or predictable model choice.

​

How model choice surfaces in Chat

Use the model selector beside the composer to choose Auto or an available model. Model settings expose supported reasoning controls. Chat can remember your last selection, while workspace defaults provide the starting choice. Higher reasoning effort can take longer. Web search, image generation, and other model tools appear only when supported by the selected route and permitted by the workspace. A model’s ability to read images does not by itself mean it can generate them. See Knowledge, tools, web search and models for the other conversation controls.

​

If a model cannot run

Check that the model is enabled, its access mode is permitted, and the selected credentials are ready. An explicit unavailable model is not a request to silently switch providers or billing modes. If access was recently changed, ask an owner or admin to review Connectors → AI and Settings → AI. NetSuite authorization is a separate requirement for NetSuite tools.