Evaluate your agent's performance by using AI-generated test cases or uploading your own test cases in a CSV template. AI-generated test cases offer two data options: generating tests based on subagents and actions, or generating tests based on the knowledge available to the agent via an Agentforce Data Library.
Required Editions
 This article applies to: |
New Testing Center in Agentforce Studio (Beta) |
 This article doesn't apply to: |
Legacy Agentforce Testing Center in Setup |
| Available in: Lightning Experience |
| Available in: Enterprise, Performance, Unlimited, and Developer Editions. Required add-on licenses vary by agent type. |
| User Permissions Needed | |
|---|---|
| To create tests in the Testing Center: |
Manage Agentforce Grids AND Manage Agentforce Testing |
How Do I Select My Test Data?
Each testing approach helps you simulate real-world interactions and catch issues early, but some are more useful than others depending on your testing goals.
Subagents and actions test cases are formatted as conversations, enabling evaluation at both the single-utterance and multi-turn conversation level. We recommend generating subagents and actions tests to evaluate how your agent handles specific intents or workflows, especially after making updates or before launching an agent. This approach checks that the agent triggers the correct behaviors based on user utterances.
Knowledge-based test cases are formatted as question-and-answer style utterances taken from the knowledge information available to the agent. We recommend generating knowledge-based tests to verify that your agent delivers accurate, content-driven responses from connected knowledge sources, such as articles or PDFs. This approach is ideal for confirming that important information, like company policies or product specifications, is correctly retrieved and communicated.
If you need more control, download the provided CSV template, fill it with your custom test scenarios, and upload it. This approach is best for tailored testing or validating specific business logic.