Anthropic and Google models are now available inside the AI Agent Action in workflows, alongside OpenAI. You can choose your provider, select a model, and control how much reasoning it applies, all from a redesigned selection interface.

What's New

Expanded Provider and Model Support

The AI Agent Action now supports models from three providers:

• Anthropic: Claude Opus 5, Claude Sonnet 5, Claude Haiku 4.5
• Google: Gemini 3.6 Flash, Gemini 3.1 Pro Preview
• OpenAI: GPT-5.6 Luna, GPT-5.6 Tera, GPT-5.6 Sol, GPT-5 Nano

Redesigned Model Picker

Models are grouped by provider in a new selection interface. Models that support reasoning are marked with a thinking chip, and each model carries a short description of what it is built for, so you can evaluate the options before committing.

Effort Control

Set reasoning effort to Low, Medium, or High from the same dropdown. Higher effort produces deeper reasoning on complex tasks; lower effort returns faster responses on simpler ones. Effort control is available on models that support thinking.

How to Use

1. Open your AI Agent Action.
2. Click the model dropdown.
3. Browse by provider, and use the thinking chip and model description to select a model.
4. Choose an effort level from the same dropdown.
5. Save and publish.

Why This Matters

A flagship model can handle multi-step reasoning where accuracy is critical, while a faster, lighter model handles intent classification, simple reads, and basic operations - allowing you to choose the right model for your task in hand.

Effort control adds a second dimension to that decision, letting you tune quality against speed and token consumption on each action independently.

Layered on top of the token optimization shipped earlier, the same task now runs up to 50% cheaper.