Agent testing · VoxVain Bench + Sim Preview

Test voice agents before they take a call.

Bench scores your agent against a suite of conversations. Sim plays realistic simulated callers, including the difficult ones, so regressions show up in a test run instead of on a real call.

Coming soonRead the docs
Today10starter caller scenarios
Target≈2 minfor a full suite run
TargetCIruns in your pipeline

“Target” figures are goals we have not measured yet, not results. Anything marked “Measured” is a number we took ourselves; your latency depends on network, region and text length.

Features

What you get.

Simulated callers

Scripted and generated callers: rushed, confused, angry, or asking for a person.

Scored runs

Pass or fail on what matters: details captured, disclosure given, no invented facts.

Regression gates

Block a release when a new prompt or voice drops the pass rate.

Latency and cost

Every run reports time to first word and cost per conversation.

Bench + Sim is coming soon.

Customers cannot run Bench yet. There is no hosted runner, dashboard or billing for it.