Prepare a Retell agent for QA

1. Prepare the test agent

A cloned test agent ID and API key. Confirm whether the agent uses a Retell LLM, conversation flow or custom LLM.

Use synthetic customer information and a dedicated test calendar. Disable access to production booking tools and live customer queues.

2. Inspect the configuration

Tapewall reads agent metadata. For a standard Retell LLM engine, it also reads the separate LLM resource and state prompts. Other engine types need an agency export.

Open Settings in the private workspace, select the platform and your cloned agent, and request a read-only inspection. Credentials stay server-side.

3. Approve the facts

Review hours, services, prices, staff and booking rules from the prompt. Add them to the fact registry and approve them before creating scenarios. Imported text is not automatically approved.

4. Check the booking, not just the answer

Agree on written test scope, connect the isolated calendar, and complete a supervised test before scheduling. A booking needs matching service, time, staff and synthetic customer evidence.

Native testing support differs by engine type. A provider test pass cannot prove an appointment exists in your booking system.

When the prompt changes

A new manual inspection compares the available prompt and version with the saved snapshot. Detected changes require review before live tests resume. This does not continuously monitor every tool or remote setting.