OpenAI
Connect OpenAI's models — GPT-4o family, GPT-4o mini, embedding models — to your workspace.
What you need
- An OpenAI API key with billing enabled. Create one in the OpenAI platform under "API keys".
- The model ID you want to use (see recommendations below).
- A few minutes.
Step by step
- Open the Models tab and click Configure new Model.
- Pick OpenAI as provider.
- Enter a display name. Suggested: include the model name, e.g. "OpenAI GPT-4o mini".
- Enter the model ID, for example
gpt-4o-mini. - Enter your API key in the secured field.
- Pick the model from the list. For embeddings, connect a second OpenAI model and pick an embedding ID (e.g.
text-embedding-3-small) via the Custom / Other option — chat and embedding are separate connections in the workspace. - Click Create. The model is saved and available in the workspace.
Recommended model IDs
The exact list shifts over time. As a general guide:
| Need | Model ID | Notes |
|---|---|---|
| Fast and inexpensive chat | gpt-4o-mini | Good default for most automations. |
| High-quality chat & reasoning | gpt-4o | More expensive; reach for it when quality matters more than cost. |
| Embeddings (default) | text-embedding-3-small | Cheap and accurate enough for most knowledge bases. |
| Embeddings (high accuracy) | text-embedding-3-large | Higher dimensions; better recall on tricky data. |
Pick what works for your use case. If OpenAI releases a newer model, just type its ID in the field — the AI Kit does not gate against unknown IDs.
Pricing
OpenAI bills per token, separately for input and output, and for embeddings per token. Current rates live on OpenAI's pricing page. Costs are billed to your account; see BYOK.
Rate limits and quotas
OpenAI sets rate limits per organization and per model. Limits scale up as your account ages and as your spending increases. If automations start failing with rate-limit errors, check your OpenAI dashboard — you may need to request a limit increase.
Recommendations
- ✅ Create a dedicated API key for the AI Kit. Label it so you can revoke it without affecting other applications.
- ✅ Connect one chat model and one embedding model to start. Add more only if a specific automation needs a different model.
- ✅ Set a monthly spend limit in the OpenAI dashboard until you have a feel for the cost.
- ⚠️ OpenAI occasionally retires older model IDs. If a model abruptly stops working, check OpenAI's deprecation notes.
- ❌ Do not use a key tied to a personal account if the workspace is for company use. The owner leaving the company breaks every automation.
What to do next
- Add another provider for comparison: Anthropic, Mistral.
- Keep data on your own infrastructure: Ollama (Self-hosted).
- Make sure you understand BYOK: Bring-your-own-Key.