Before your agent meets its first users.

Work through the checks that apply to your integration. These checkboxes stay in this page only; they do not run tests or send results anywhere.

Project and permissions

The application uses the intended project and assigned line.Test users meet the plan’s verification or allowlist rules.Provider credentials stay on the server and are scoped to this environment.The workflow fits the plan’s outbound permissions, minutes, rate limits, and concurrency.

Messaging — if your application uses it

A permitted inbound message reaches the correct application user and conversation.A repeated event does not trigger the same business action twice.A worker restart resumes persisted work without losing an acknowledged event.Delivery outcomes and errors are observed separately from API request acceptance.

FaceTime — if your application uses it

The caller hears the agent and the agent responds to the caller’s actual words.Required agent video changes during the call; a still image does not count.If caller vision is required, the agent responds to a scene change not described in the prompt.Outbound greetings wait for bridge readiness.Hangup releases resources and a second call starts a fresh conversation.

Operational readiness

Busy lines, missed calls, provider failures, and expired tasks have a defined fallback.Consequential tool actions require the appropriate user authorization.Analytics measure a completed useful task without collecting unnecessary conversation content.The team knows where to find sanitized errors and which release is deployed.