Basic report
Free
20 test conversations
- A sample of applicable checks
- Up to 3 key findings
- Selected test transcripts
- Basic report delivered by email
ACTOUT AGENT CHECK
We test your AI agent’s answers and actions.
Send us a link and tell us what your agent should do. We run multi-turn conversations, compare its behavior with your rules and report failures you can reproduce.
You get the test conversation, the expected result and what actually happened, so your team can see where it went wrong.
How we run the test
Agree on expected behavior
We confirm access and define test cases from your agent’s purpose, business rules and reference material. Each case has an expected result.
Run conversation scenarios
We vary wording, add follow-up questions, change requirements and try adversarial prompts. We record the responses and rerun suspected failures.
Review the evidence
We compare observed results with the expected behavior and report the failures. The detailed report adds severity, suggested fixes and a focused retest.
A public link lets us test the conversation. Verifying tool calls, access controls or backend changes requires a test environment or logs. Anything we cannot verify is marked as not tested.
Choose the level of testing you need
One agent · One language · Up to 3 tasksDeep agent test
$149 one-time
100 test conversations
- All applicable checks
- All findings, prioritized by severity
- Full transcripts and suggested fixes
- Critical findings reviewed by our team
- 1 focused retest
We confirm access before testing. For private agents, APIs or voice interfaces, contact us to agree on scope.
Find out how your agent handles the conversations that matter
Share your agent’s link and the work it should do. We’ll test its responses and actions, then show you what needs attention with evidence your team can use.