> For the complete documentation index, see [llms.txt](https://docs.meetalto.ai/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.meetalto.ai/using-meet-alto/agents/testing.md).

# Test before going live

Use preview conversations and simulations to check accuracy, actions, and handoffs before launch.

Use **Preview** in the agent editor to try a conversation and **Test** to open the simulation checks. Save the agent first if the test control asks you to. Follow the readiness messages shown for your saved version.

## Try a useful range of conversations

| Case                                     | What to check                                   |
| ---------------------------------------- | ----------------------------------------------- |
| A common question                        | Correct facts and a direct answer               |
| A vague request                          | One useful clarification instead of a guess     |
| Missing knowledge                        | Honest uncertainty and a practical next step    |
| An interested customer                   | The appropriate approved action or link         |
| A customer who declines                  | No repeated pressure or irrelevant follow-up    |
| A complaint or exception                 | A clear route to your team                      |
| Conflicting instructions from a customer | Business rules and approved facts remain intact |

Use fictional customer information. Test a whole exchange, including follow-up questions; a good first reply does not prove the rest of the conversation works.

## Read the result carefully

Inspect the conversation and individual findings. **Couldn't run** is different from **Passed**. Resolve the cause of an incomplete test rather than treating an absent failure as success.

If you edit the playbook, save and rerun the relevant cases. Check that the test result applies to the saved version. Testing allowances may apply; use the quota shown in the test panel rather than repeatedly running an unchanged scenario.

Preview quality does not prove that a real provider connection can deliver a message or perform an external action. Start live use with **Training wheels**, inspect the Inbox, and verify any destination with a test contact you control.

Continue with the [launch checklist](/using-meet-alto/playbooks/launch-checklist.md).