Honeybase AI credits

What it is

Every workspace registered through the sign-up flow starts with 200 Honeybase AI credits — one credit is one US cent of model usage, so the grant is worth about two dollars of calls. The credits are usable the moment the workspace exists: pick the "Honeybase AI credits" connection wherever a model is chosen — Barry's composer, an agent's settings, an AI node — and it works with no API key and no setup.

Behind it, Honeybase runs the calls on its own provider account through a per-workspace budget, so one workspace can never spend another's credits and the grant is the hard ceiling. The model is fixed by Honeybase (currently openai/gpt-5.6-sol) and may change; the connection's model list shows what is available.

Not set up through a connection form

There are no connection fields to fill in. Honeybase creates this connection when the workspace is registered, holds the credential, and routes the calls — which is why it appears on the Integrations page but cannot be added, edited or removed from Add integration.

When the credits run out

Calls on this connection stop, deterministically, when the grant is spent. Barry, agents and AI nodes say so rather than surfacing a provider error. From there there are two exits: bring your own key with an AI connection and switch the affected agents and nodes to it, or buy more credits (not yet available).

Credits vs. bring your own key

The credits connection is a way to start, not a replacement for your own provider account. An AI connection with your key has no Honeybase-side budget, lets you choose the provider and model, and keeps usage on your own bill. A workspace can have both; each agent and node picks which one it uses.

Privacy

Under an AI connection you bring, the provider is your choice. Under Honeybase AI credits, prompt content goes to the provider Honeybase has chosen for the credits route; see the privacy notice for the current sub-processors. The route is configured to opt out of provider-side data collection.