Testing, Deployment, and Maintenance10% of the exam
Testing agents in Testing Center
How to batch test an agent in Testing Center: why it runs in a sandbox, who can use it, turn-level versus conversation-level suites, generating or uploading test cases, the CSV template's API-name rules, and the limits that shape a large test plan.
Given a scenario, test an agent using Testing Center.
~8 min read · 9 practice questions ·
What the exam expects
- Choose Testing Center batch testing over manual Conversation Preview testing, and say what each is for.
- Pick the right environment and permissions for a test run.
- Choose how to create test cases: generated from subagents and actions, generated from knowledge, or uploaded as a CSV.
- Fill the CSV template correctly, including API names for expected subagents and actions.
- Plan around the limits of large suites and of conversation-level (multi-turn) testing.
- Key terms: utterance (one user input sent to the agent); test case (an utterance plus its expected outcomes); test suite (a named set of test cases run against one agent version); turn-level test (scores one response); conversation-level test (scores a whole simulated multi-turn conversation); subagent (formerly topic) (the job category the agent routes an utterance to).
Key facts
Where to run tests
- Detail
- Sandbox only; tests can modify CRM data
Studio (Beta) permissions
- Detail
- Manage Agentforce Grids AND Manage Agentforce Testing
Legacy Setup permissions
- Detail
- Manage AI Agents + agent-type permissions + System Admin profile (or Customize Application + System Admin)
Version tested
- Detail
- Must be committed; defaults to latest
AI generation per request
- Detail
- Up to 200 test cases (Studio)
Knowledge-generated tests
- Detail
- Turn-level only; articles, knowledge fields or uploaded files
CSV required column
- Detail
- Utterance (turn-level) or Conversation (conversation-level)
Expected subagent and actions
- Detail
- API names, actions as a Python-style list
Size and concurrency
- Detail
- 1,000 tests per uploaded file; Trailhead: 10 test jobs at once per 10 hours
Test cost
- Detail
- Unmetered since Summer '26, but Data 360 queries stay metered; older Trailhead says tests consume credits
Conversation-level beta limits
- Detail
- 3 concurrent suites, 20 turns; no context variables, custom scorers or Employee Agents
Sources for this section (9)
- Agentforce Testing Centerhelp.salesforce.com (opens in a new tab)
- Set Up and Run Tests in Agentforce Studio (Beta)help.salesforce.com (opens in a new tab)
- Effective Testing Setup for Agentforce Agentstrailhead.salesforce.com (opens in a new tab)
- Create a Test Suitehelp.salesforce.com (opens in a new tab)
- Generate Tests from Subagents and Actionshelp.salesforce.com (opens in a new tab)
- Generate Tests from Knowledge Sourceshelp.salesforce.com (opens in a new tab)
- Upload Test Caseshelp.salesforce.com (opens in a new tab)
- Understanding Trust in Agentforce Testing Methodstrailhead.salesforce.com (opens in a new tab)
- Considerations for Testing Centerhelp.salesforce.com (opens in a new tab)
- Why must Testing Center run in a sandbox?
- When do you generate tests from knowledge sources instead of from subagents and actions?
- Why does a correctly routed test fail when Expected Subagent holds the builder label?
- Which two features does conversation-level testing not support that turn-level does?
- How many tests can one uploaded CSV file hold?
The questions are free to read. Answers with their sources and self-grading are included with the pack.
Keep reading with the Agentforce Specialist pack
- Core concepts
- Exam traps
- Worked scenario
The pack includes:
- All 26 lessons of the Agentforce Specialist guide, with reading progress
- Interactive decision walks and step-by-step sequence steppers
- Inline practice questions and one-click topic practice (9 questions for this lesson)
- An official source cited for every claim
- An AI tutor that answers from the guide
- Unlimited timed practice exams