In the new Agentforce Builder, Canvas view now includes powerful features that were previously only available in Script view. We also introduced new features and improvements that streamline the agent development process.
Where: This change applies to Lightning Experience in Enterprise, Performance, Unlimited, and Developer Editions with Foundations or Agentforce 1 Editions.
Note The new Agentforce Builder is a beta service that is subject to the Beta Services
Terms at Agreements - Salesforce.com or your written Unified Pilot
Agreement, and applicable terms in the Product Terms Directory. Use of this beta
service consumes usage types per the Agentforce and Generative AI Usage and Billing
documentation and is at the Customer’s sole discretion.
When: This change is available starting the week of January 5, 2026.
Why:
- When you add an action to a topic, the action is automatically added to the topic’s reasoning actions. Previously, you had to manually reference the action in your topic before the agent could use it.
- Create, add, and manage variables on the Variables page in Canvas view. Also, you can create a variable that's set to an action output in a topic’s actions available for reasoning (reasoning actions). Previously, you could only create, add, and manage variables in Script view.
- Add filters in the Topic Selector and in a topic’s reasoning actions. When a topic or action has a filter, your agent can use it only when the filter conditions are met. Previously, you could only create filters in Script view.
- Add follow-up actions in a topic’s reasoning actions. Follow-up actions let you tell an agent what to do immediately after an action runs. For example, you can have the agent chain actions or transition to a specific topic. Previously, you could only define follow-up actions in Script view.
- Add agent-level instructions on the System Messages page in Canvas view. Agent-level instructions apply to all of an agent’s reasoning and interactions. You can use them to define an agent’s tone, persona, style, and goals.
- The Filter from agent action setting on agent action outputs has been renamed to Include in agent context, to better reflect its functionality in the new Agentforce Builder. When the setting is enabled, the output is added to the agent’s conversation history and made available for reasoning. Disable the setting for any underlying reference action outputs you don’t want to use with your agent.