Catch what breaks before callers do.
UK-accented callers ring your phone agent over phone-quality audio (8 kHz), with noise and interruptions. Every call is scored, with the line behind the score.
Works with any agent you can ring, built on Vapi, Retell, LiveKit, Pipecat, Twilio or your own stack.
- test calls
- 240
- passed
- 87.1%
- median reply
- 0.88 s
Reads policy numbers back wrong after a correction.
One run, start to finish
A first run of 60 calls on a made-up claims line.
Calls go out
Simulated callers ring your agent. Each dot is one call.
Every call is scored
AI judges mark each call against the pass mark, quoting the line that decided it.
Failures group by cause
Twelve failed calls become three things to fix, ranked by risk.
You get the evidence
Each finding comes with the quote, the score and the fix.
Top finding: reads the number back wrong
This is high risk. The caller says the policy number with a correction, and the agent reads back a different number. 80% of calls passed. The fix is to read the number back in groups and wait for a yes.
What each run covers
- Your own scenariosDescribe a caller in a sentence and we add them to the run.
- Scored per ruleEach call gets a pass or fail for every rule you set.
- Grouped by causeFailed calls become a short list of fixes, ranked by risk.
- Release checksRun the same calls after every change to your agent.
- Nothing to installGive us a Vapi or Retell agent ID and API key, or a phone number.
- API key deleted after the runWe encrypt it, use it for that run only, then delete it.
Your first 20 calls are free.
You get one report, with an accent caller and a noisy line, and no card is needed.
Six kinds of test caller
Made-up callers on a phone-quality line.
“It’s the flat above the chippy, number 4B.”
“Thursday’s fine. Actually, no, can we make it Friday?”
“Er, 22 Station Road?”
What built-in platform tests miss
Most voice platforms ship their own test tools, and they are a good place to start. Vapi’s docs note what they leave out:
“Use manual calls and production recordings to test background noise, degraded audio, a range of accents, silences, and interruptions that your Simulations cannot reliably reproduce.”
From Vapi’s test suites documentation, read 28 October 2026.
That is the gap RingTru is built for. Our made-up callers have UK accents, ring over phone-quality audio (8 kHz) with background noise, and cut in or change their minds mid-call. Every call is scored against your rules, so you do not have to make those calls by hand.
Accents and noisy linesEvery score shows its evidence
Disagree with a mark and you can overturn it, with a note saying why. We will publish how often judges agree with people.
How the judges workRule: before booking, read the day back and wait for a yes.
CallerThursday’s fine. Actually, no, can we make it Friday?
AgentLovely, the assessor is booked for Thursday.
The caller asked for Friday. The booking was made for Thursday and not read back.
Learn how agents fail
Guides for the people who build and run voice agents. All guides
Plans in US dollars
Prices exclude tax. Compare plans
How we handle your data
The promises we build to. Read them in full
- Keys deleted after the runAn API key is encrypted, used for that run only, then deleted.
- Made-up callers onlyWe never use a real person’s voice, name or details in a call.
- No training on your callsRecordings and transcripts are used to score your run and nothing else.
- No cookies, no analyticsThis site sets no cookies and runs no analytics. Privacy notice
Run a test on your own agent.
RingTru opens soon. Join the waitlist and your first 20 test calls are free.
To join the waitlist, email us at greg@gme.solutions.
Rather talk first? Book a call
- test calls
- 506
- passed
- 96.8%
- median reply
- 0.94 s
Holds the new day when a caller changes their mind.