Tines 3B runs on the AI models you bring to it. Before you can build and run workflows, you need at least one AI provider connected and set as the default. This guide walks you through adding a provider in settings, pointing it at the right models, and making it the one Tines 3B uses.
Note: Connecting a provider is a tenant admin task. You'll need the manage AI providers permission on your tenant. If you don't have it, the AI providers screen shows "Only admins can change this," and you'll need someone with admin access to make the change.
You'll also need a connector that holds your provider's API key.
In Tines 3B, API keys are never typed directly into the AI provider form. They come from a connector, which keeps the secret stored securely on the server. You can pick an existing connector or create a new one while you set things up.
To set one up, go to the AI providers section in your settings, like this:
You'll see a table of any providers already connected, along with their chat model, fast model, default effort, and whether each one is enabled.
If your deployment includes Tines 3B managed models, you'll see a "Managed by 3B" row covering Claude and OpenAI, which you can use without connecting anything yourself.
Choose a provider to connect
Select the plus button. You can connect any of these:
OpenAI. Connect directly to OpenAI's API.
Anthropic. Connect directly to Anthropic's API.
Azure OpenAI. Connect through Azure OpenAI, using a base URL and deployment names.
Oracle Cloud Infrastructure. Connect to OCI Generative AI, with requests signed by a connector.
Custom. Connect to any OpenAI-compatible API endpoint.
Selecting a provider opens the setup modal. The fields you see depend on the provider you picked. A few appear for most providers:
Base URL. For OpenAI and Anthropic, the endpoint comes from your connector, so there's no separate base URL field. For Azure OpenAI and Custom, a base URL is required. For OpenAI-compatible endpoints, use your provider's API base URL.
Region. Oracle Cloud Infrastructure asks for your OCI region slug, for example "us-chicago-1."
CA certificate. The Custom provider lets you optionally paste the PEM CA certificate your endpoint uses.
You won't find an API key field here. The key lives in the connector you attach in the next step.
Attach a connector
The connector field is where your provider's credentials come from. You can either:
Choose an existing connector from the list. Tines 3B shows connectors whose URL pattern matches the provider's endpoint.
Select New connector to create one inline for this provider.
Once a connector is attached, Tines 3B uses it for the API key, and for OpenAI and Anthropic it also uses the connector's base URL as the endpoint.
Pick your models and effort
With the connection in place, tell Tines 3B which models to use:
Chat model. This is the main model for your workflows. Leave it blank to use the provider's default, unless you chose Custom or Oracle Cloud Infrastructure, where a chat model is required. If your connector limits the available models, you'll choose from that list.
Fast model. A lighter model for quicker tasks. Leave it blank to reuse your chat model.
Effort. For reasoning models that support it, this sets the default effort level. Users can override it per chat, so this is just the starting point.
Save and set as the default
When you save your changes, they are validated first, so if a required field is missing, saving stays disabled until you fill it in.
Tines 3B uses one provider as the default for new work. To choose it, open the menu next to the provider you want and select set as default.
What's next
Once a default provider is connected and enabled, you're ready to start building workflows in Tines 3B. If you connect more than one provider, you can switch the default at any time from this same screen, and each provider keeps its own models, effort, and pricing settings.
Learn how to start building your first workflow here.
Managing AI providers
Switching the default provider, changing the provider’s models, setting the default effort, turning providers on or off and more can all be done within the AI provider’s settings.
Change the model or effort for a single chat
Sometimes you want a different model or effort level for just one conversation, without changing the tenant default. You can do this right from the composer, and it doesn't require admin access.
In the chat composer, open the model picker. It shows the current model and provider.
Choose any enabled model from the list. The one marked Default is your tenant's default.
To adjust effort, select the effort button next to the picker. Each press cycles to the next level the model supports.
These choices apply to the current chat only. Your tenant's default provider, models, and effort stay exactly as the admin set them.