Run hundreds of calls at once
Describe a scenario in plain English. Rinwave creates versions of it with different voices, accents, background noise and caller behavior, and places the calls in parallel.
See which calls failed and where
Each call is checked for whether the agent finished the task, made up a fact, talked over the caller or paused too long. Every result links to the recording and the exact line in the transcript.
Rerun the tests whenever you change the agent
Run your saved tests on every release and score a sample of live calls. If the pass rate drops or response times go up, you get an alert.
Keep scenarios in your repo
Scenarios are short YAML files you review like any other change. A GitHub Action runs them on every pull request.
# What the caller wants, and what counts as a pass goal: Move Thursday’s appointment to Friday caller: voice: marcus behaviors: [interrupts, changes-mind, speaks-fast] variations: 500 pass_when: - booking.date == "friday" - no_made_up_facts - p95_latency < 1.5s
- name: Test voice agent uses: rinwave/test-action@v1 with: suite: reschedule min-pass-rate: 0.95 env: RINWAVE_API_KEY: ${{ secrets.RINWAVE_API_KEY }}
What's included
Engineers can run tests from their release pipeline. Product and support teams can read the reports without help.
Regression suites
Save scenarios once and rerun them on every prompt, model or tool change.
Accents and languages
Test with the accents and languages your customers use.
Interruption testing
Callers talk over the agent mid-sentence, including while it reads back details.
Response-time tracking
See how long callers wait for each reply, per turn and per scenario.
CI and release gates
Trigger tests from your pipeline and stop releases that score below your pass rate.
Replays and transcripts
Go from any failed check to that point in the recording.
Run your first 100 test calls free.
Connect your agent by phone number or API. No card needed.