SoonVoice QA stress-tests agents under real call conditions

Test your voice agents before your customers do

Brook calls the numbers your AI agents answer, asks real-world questions, and tells you whether they're good, working, or failed, before a customer ever finds out.

No SDK or code changes Free for 1 agent Live in 2 minutes
Brook · test call
live
Sales IVR
+1 (415) 555-0142 · Twilio · US-West
Dialing agent…
Scoring call…
Listening for answer, latency, and intent match.
test 1 of 3
How it works

From phone number to verdict in three steps

Brook behaves like your toughest customer — on a schedule, around the clock — and turns every call into a clear pass or fail.

01
Add your numbers
Point Brook at the phone numbers your voice agents answer. Twilio, Vonage, or any carrier. No SDK, no code changes.
02
We place real calls
On your schedule, Brook dials each agent and asks real-world questions out loud, then listens to the spoken reply.
03
Get a clear verdict
Every call is scored Good, Working, or Failed, with latency, transcript, and the exact question that tripped it up.
Features

Everything you need to trust your voice agents

Built on the same monitoring engine as the Brook dashboard — real calls, real transcripts, real verdicts.

Real spoken calls
Not a ping or a healthcheck. Brook talks to your agent the way a customer would: voice in, voice out.
Answer scoring
Match spoken replies against expected answers to catch wrong, vague, or off-policy responses.
Latency tracking
p50 / p95 response time on every call so you know when an agent is slow before customers feel it.
Full transcripts
Read exactly what was asked and said on every test, with the audio attached.
Smart alert routing
Route failures to email, Slack, Discord, or a custom HTTP endpoint for your own backend, only when it matters.
Public status pages
Share live uptime for your voice agents with a branded status page.
Scheduled monitoring
Run checks every minute or every hour, around the clock, across regions.
Regression catching
Ship a new prompt and instantly see if comprehension dropped against your test suite.
The verdict

Three answers. Zero guessing.

Every test call resolves to one of three states — the same language as your dashboard's Operational, Degraded, and Down. No dashboards to decode, no log-diving. Just: is my agent good for customers right now?

Good
Answered fast, understood the question, and gave a correct, on-policy reply. Customers get the right answer.
Working
Picks up and responds, but it's slow, asks for repeats, or partially misses intent. Degraded. Fix before it slips.
Failed
No answer, dead air, or a wrong answer to a real question. The agent is down for customers right now.
1.4M
Test calls placed every month across customer agents
30s
Median time to detect a down voice agent
99.98%
Monitoring uptime, we watch the watchers
4
Continents with local dialing points of presence
On the roadmap

From health checks to full voice QA

Today Brook tells you when an agent breaks. Next, we're building the tools to break it first — and to run all of it wherever you need.

Up next
Voice QA
Stress-test your agents under the conditions real callers create: noisy backgrounds, choppy audio, accents, interruptions, and irate customers. Brook finds the edge cases where your AI breaks before they reach production.
Planned
Self-hosting
Run health checks and Voice QA entirely on your own infrastructure, for offline, air-gapped, or compliance-bound environments that can't send calls to the cloud.
Planned
Enterprise support
Dedicated support, onboarding, and SLAs for teams running voice agents at scale across many numbers and regions.
Pricing

Pay for the numbers you watch

Every tier places real spoken test calls against your live agents. Move up only when you add numbers, never to unlock the basics.

Free
For a single agent
$0/mo
no card required
  • 1 monitored number
  • 100 test calls / mo
  • Hourly scheduling
  • Email alerts
  • 7-day history
Most popular
Pro
For growing voice teams
$63/mo$79
billed annually
  • 25 monitored numbers
  • 2,500 test calls / mo
  • 1-minute scheduling
  • Slack, Discord & custom HTTP endpoints
  • Answer scoring & transcripts
  • Public status pages
Enterprise
Scale
For agencies managing client numbers
Custom
volume-based · billed annually
Number tiers
1002505001,000+
  • Tiered by number volume, not per-seat
  • Volume test-call pricing
  • Multi-region dialing
  • SSO, SAML & audit logs
  • Custom test suites & SLAs
  • Dedicated support & onboarding
Included on every plan
  • Real spoken test calls
  • Pass / fail verdicts
  • Works with any carrier
  • No SDK or code changes
FAQ

Questions, answered

Brook places a real outbound phone call to the number your voice agent answers, speaks a real-world question using text-to-speech, then transcribes and scores the spoken reply. It's the same path a customer takes — no SDK or code change required.
Yes. If your agent answers a phone number, Brook can dial it. We support major carriers out of the box and place calls from points of presence across four continents for realistic latency.
Failed means the agent didn't answer or gave a wrong answer to a real question; customers are affected right now. Working means it responded but was slow, asked for repeats, or partially missed intent. Good means it answered fast and correctly.
On Pro and Scale you define test suites: the exact questions to ask and the answers to expect. Brook scores every call against them, so you catch regressions the moment a prompt change drops comprehension.
Median detection time is around 30 seconds. Failures route instantly to email, Slack, Discord, or a custom HTTP endpoint, and only fire when a real check fails, so you're not drowning in noise.
Voice QA is up next: it stress-tests your agents under real call conditions like noisy backgrounds, choppy audio, and irate callers to surface the edge cases where they break. Self-hosting and dedicated enterprise support are also on the roadmap for teams that need offline checks or run voice at scale.
Yes: one monitored number, 100 test calls a month, and hourly checks, free forever with no card required. It's enough to keep a single production agent honest.
Start free

Hear what your customers hear

Add your first number and Brook places a real test call in minutes. No SDK, no credit card.

  • 100 free test calls every month, forever
  • Works with Twilio, Vonage & any carrier
  • Email, Slack, Discord & custom HTTP alerts the moment something fails
SOC 2 Type II in progress · we never store call audio without consent