Skip to content
Talk to an engineer

Agent Studio

AI Provider Keys

Path: Settings → AI Gateway → AI Provider Keys

AI Provider Keys are the LLM provider credentials your agents and task agents run on: your own OpenAI or Gemini key, or the built-in Foldspace key. Foldspace uses them to call the model on your behalf. A built-in Foldspace key ships by default, so everything works out of the box, and you can add your own key from a supported provider.

Each row is a key, with:

ColumnWhat it shows
Name / DescriptionYour label for the key.
ProviderThe key’s provider (Foldspace built-in, OpenAI, or Gemini).
UsageThe agents and task agents currently using this key. Open it to see the full list.
DefaultA star marks the default key.
Created By / Created DateProvenance.

Foldspace includes a built-in Foldspace key (the “Default Foldspace Key”) so new accounts have a working model immediately.

  1. Click Create.

  2. Fill in the key:

    FieldPurpose
    ProviderThe LLM provider: OpenAI or Gemini.
    NameA label for the key.
    DescriptionOptional note.
    Secret KeyThe provider’s secret API key.
    DefaultWhen on, any new Agent or Task Agent automatically uses this key.
  3. Click Test Connection to verify the key reaches the provider, then Create.

Failover can only route to keys you’ve added. Add a key on both providers, one OpenAI and one Gemini, so that when one provider is throttled or down, Foldspace can cross to the other on your own key instead of falling straight back to the built-in Foldspace key. Adding a second key for the same provider also lets failover spread load across your keys. See High Availability for the full fallback order.

The starred key is the default: new agents and task agents automatically use it. You can change which key is default at any time. To see exactly which agents and task agents a key powers, open its Usage: it lists every Agent and Task Agent (with links) bound to that key.

  • High Availability: how failover routes across your provider keys.
  • Token Quotas: cap token spend per agent, user, and tier.
  • User Management: the AI Provider Key feature is one of the permissions you can grant per teammate.
  • Cost & tokens: track spend and token consumption per model and provider.