Today, we are excited to announce Agent Script Recipes, a new sample app designed to help you master hybrid reasoning. Moving from deterministic code to probabilistic AI agents often feels like a leap. Agent Script Recipes bridges that gap for you. With this initial release, we’ve provided more than 20 clear, bite-sized examples to copy, paste, and adapt.

If you develop intelligent agents that combine structured code with flexible AI, then this post is for you, especially if you care about long-term maintainability. We created this collection of Agent Script examples to demystify the language and help you quickly implement solutions with Agentforce.

Screenshot of an Agent Script view within Agentforce Studio

Explore key hybrid reasoning patterns

The core value of this sample app is the ability to see logic flow between your code and the Atlas Reasoning Engine, and to understand the core patterns of the language. Instead of starting from scratch, you can explore a working repository that highlights specific implementation details.

We organized the app into specific recipe patterns to solve common development challenges, including:

  • Language Essentials: These recipes cover the basic language fundamentals that everyone should be familiar with, like how to work with variables in Agent Script, or how to use template expressions.
  • Action Configuration: Action recipes focus on demonstrating on how to call specific actions with Agent Script, like Apex, flows, or prompt templates.
  • Reasoning Mechanics: A key capability of Agent Script is the ability to add deterministic control when you need it, or leave control to the large language model (LLM).This is demonstrated in the reasoning recipes.
  • Architectural Patterns: For more advanced examples, you can explore the architectural recipes. These showcase, for example, more complex state management, or how to orchestrate multiple agent topics.

      See Agent Script Recipes in action

      A picture is worth a thousand words. And a video even more. Check out the latest video from our Agentforce Decoded playlist, which walks you through how to get started with the sample app.

      Learn implementation details for Agent Script Recipes

      Moving from building probabilistic to more deterministic agents requires a new way of thinking. As we were building out Agent Script Recipes, we added a dedicated README for every agent, which summarizes the agent flow, key concepts, or try-it-out utterances.

      And to make the README files more accessible for everyone, we also provide them on the Salesforce Developers website.

      Screenshot of Agent Script Recipes documentation on developer.salesforce.com

      Next steps

      Hopefully, you now have a better understanding of how the Agent Script Recipes sample app can accelerate your Agentforce development. To get started, clone the repository, explore the recipes, and deploy them to a Developer Edition or scratch org.

      We look forward to seeing the powerful agents you build. Don’t forget to share your feedback on the Trailblazer Community, LinkedIn, or X using #AgentScript.

      Be sure to tune into two Agent Script virtual sessions this month:

      About the author

      René Winkelmeyer leads the Developer Advocacy team at Salesforce. His team focuses on Agentforce, Data 360, MuleSoft, and the Salesforce Platform. In his spare time, you might find him still coding on GitHub. Follow René on LinkedIn.