AI registry connections

Overview

Each AI registry in the catalog maps to a provider (e.g. OpenAI, Gemini, Runway, Anthropic). Your team stores API keys as one or more named connections per provider. Connections for a provider are shared by every registry that uses that provider. Some paid plans also allow default (platform) keys for certain registries; then an AI model task can omit a connection.

Where to manage connections

1. AI Models – Open AI Models, choose a registry card, then use API connections on the registry detail page to add, edit, or delete connections.
2. Workflow builder – In an AI model task, use the API connection field and the link to the registry page.

Ensure the correct team is selected; connections are per team.

Providers

Supported providers include OpenAI, Gemini, Runway, Anthropic, Mistral, Cohere, Groq, Grok (xAI), Replicate, Fal, and Stability AI, depending on which registries are enabled for your workspace.

How to add or update a connection

1. Go to the registry detail page for any registry that uses that provider (or use the link from the workflow task).
2. Click Add connection, enter a name and API key, and save.
3. To rotate a key, edit the connection and enter a new API key.

How connections are used at run time

If the task selects a connection, that key is used. Otherwise, if your plan allows a default key for that registry, the platform key may be used. If neither applies, the run fails until you add a connection or upgrade the plan.

Security

API keys are stored encrypted. Full keys are not returned in list APIs. Team roles owner, admin, and developer can manage connections.