# Every failure mode, found before launch.

Run your agent against hundreds of simulated callers: realistic personas, adversarial red teaming, native accents in 45 languages, peak-volume load and always-on health checks. Ship knowing exactly what passes.

Start free with $50 in credit, no card needed. Works with Vapi, Retell, LiveKit, Pipecat + your stack.

## Simulate before launch

### Break it in staging, not in production.
Run your agent against hundreds of simulated callers (realistic personas, accents, background noise and edge cases) and get every conversation scored before a customer ever dials in.

### Scenarios & personas
Hundreds of simulated callers (the angry one, the rambler, the interrupter) built from your real call types.

### Red teaming
Adversarial callers that try to break it (prompt injection, jailbreaks, social engineering) so your agent holds policy under attack.

### 45 languages & accents
Native accents, code-switching and background noise, in every market your agent answers.

### Load & health tests
Peak-volume concurrency and always-on health checks, so the agent that passed in staging survives launch day.

### Regression testing
Rerun the whole suite on every change and diff it against your last green baseline, so fixing one caller never breaks another.

### Run it in CI
Every prompt or model change runs the suite before it merges: quality gates for conversations, not just code.

Pre-launch suite · booking_v2182 / 200 passed

- Angry caller · refund demand: pass · 92
- Red team · prompt injection: pass · 90
- Gulf Arabic · lobby noise: pass · 88
- Interrupts mid-disclosure: fail · 61
- Rambler · 3 intents in one turn: pass · 85
- Peak load · 250 concurrent: pass  
1 failure filed as an issue: fix it before launch, not after

## Load & health

### Survives launch day. And every day after.
Passing in staging is not the same as holding at peak volume, and holding at launch is not the same as being up next Tuesday. Roark tests both, on a schedule.

### Load testing
Ramp hundreds of concurrent simulated callers against your agent and watch latency, error rate and quality under pressure, so the peak-volume Monday your forecast promised is a rerun, not a surprise.

- Load test · booking_v2: ramp to 250 concurrent
  - Time-to-first-word · p95: 680ms
  - Turn latency · p95: 1.4s
  - Dropped or errored calls: 0.2%  
held at 250 concurrent for 20 minutes · passed

### Health checks
Always-on probes call your agent around the clock and verify it answers, speaks and behaves. When one fails, your team hears it from Roark, not from a customer.

- Health check · every 5 min: uptime 99.98%
  - 09:05: Answered · first word in 1.1s: pass
  - 09:10: Answered · first word in 1.2s: pass
  - 09:15: No answer after 30s: fail  
Alert fired: #voice-alerts pinged in under a minute.

### Built for 45 languages
#### The way the world actually calls.
Native accents, code-switching and local noise, not English with a filter. Every audio-native metric runs in each one.

- 45 languages & dialects, growing every release:
  - Latin American Spanish
  - Gulf Arabic
  - Brazilian Portuguese
  - South Indian English
  - UK English
  - Australian English
  - Parisian French
  - Mandarin
  - Cantonese
  - High German
  - Japanese
  - Tagalog
  - Hindi
  - + 32 more

## Get started

### First suite running in minutes.
Pick scenarios, personas and metrics, point the plan at your agent's endpoint and launch it: from the dashboard, or a few lines of code.

```javascript
import Roark from '@roarkanalytics/sdk'

const client = new Roark({ bearerToken })
await client.simulationRunPlan.create({
  name: 'booking_v2 · pre-launch',
  direction: 'INBOUND',
  scenarios, personas, metrics,
  agentEndpoints: [{ id: agent.id }],
  maxSimulationDurationSeconds: 300,
  autoRun: true, // suite starts now
})
```

Node · Python, plus a REST API for CI/CD and webhooks the moment the suite finishes.

### Security details:
SOC 2 Type II! HIPAABAA available

Enterprise-grade from day one: annual pen tests, SSO/SAML, role-based access, configurable retention.

## Bring a recording. We’ll score it live.
See your own agent measured on the audio it actually produced, in the demo, in real time. Stop guessing whether your voice AI works.