Welcome to Roark - Roark Docs

Roark is where voice and chat agents prove they’re ready. Simulate your agent against the conversations that matter before launch, monitor every production call in real time, and score them all with metrics powered by Roark Prism — our purpose-built evaluation model. Catch what breaks, prove the fix, ship with evidence.

Monitor

Every call is transcribed, analyzed for sentiment, emotions, and speech patterns, and made searchable in real time.

Measure

Collectors run your metrics on every matching conversation — from response time to custom LLM judges — with pass/fail thresholds.

Simulate

Test agents with synthetic callers across customer flows, personas, accents, and edge cases before deploying.

How It Works

Get calls into Roark

Connect a voice platform (Vapi, Retell, ElevenLabs, LiveKit, others) and calls sync automatically. Or upload recordings directly via the SDK.

Calls are analyzed automatically

Every call is transcribed, then run through Roark’s voice analysis models — detecting sentiment, 64+ emotions, interruptions, speech pauses, and vocal cues. Active collectors run any additional metrics you’ve configured.

Define what good looks like

Use built-in metrics like response time, or author custom LLM-judge metrics in Studio (e.g., “Did the agent verify the caller’s identity?”). Add thresholds to turn values into pass/fail outcomes.

Test before you ship

Run simulations against your agent with synthetic callers. Attach customer flows and pass/fail checks to your plan — if the agent doesn’t meet your bar, you know before customers do.

Quick Start

Upload a Call
Run Metrics on a Call
Connect a Platform

Send a call recording and let Roark handle the rest:

import Roark from '@roarkanalytics/sdk'

const client = new Roark({
  bearerToken: process.env.ROARK_API_BEARER_TOKEN,
})

const call = await client.call.create({
  recordingUrl: 'https://example.com/recording.wav',
  startedAt: '2025-01-15T10:00:00Z',
  interfaceType: 'PHONE',
  callDirection: 'INBOUND',
  agent: { name: 'Support Agent' },
  customer: { phoneNumberE164: '+15551234567' },
})

The call appears in Call History with full transcription, sentiment, and emotion analysis. Any active collectors run automatically.

Collect specific metrics against existing calls:

// Create a custom metric
const metric = await client.metric.createDefinition({
  name: 'Identity Verified',
  outputType: 'BOOLEAN',
  analysisPackageId: 'your-package-id',
  llmPrompt: 'Did the agent verify the caller identity before proceeding?',
})

// Run it against calls
const job = await client.metricCollectionJob.create({
  callIds: ['call-id-1', 'call-id-2'],
  metrics: [{ id: metric.data.id }],
})

// Fetch results
const results = await client.call.listMetrics('call-id-1')

Or author and test metrics interactively in Studio first.

Explore the Docs

Observability

Call history, traces, reports, and dashboards

Metrics

The metric library, Studio, collectors, and thresholds

Simulations

Customer flows, personas, templates, plans, and schedules

Integrations

Vapi, Retell, ElevenLabs, Leaping, LiveKit, Pipecat, and custom

SDKs

Node.js, Python, and MCP Server

API Reference

REST API documentation