Barge-in & overlap QA for voice agents

A $499, one time barge-in test report for your voice agent, back this week.

Send us recorded calls, get back a scored report on barge-in, overlap, and interruption recovery. One time price, no platform to onboard, no subscription. Built for teams running Vapi, Retell, Bland, Pipecat, or self-hosted LiveKit.

Get your report $499 for the first 10 companies, then $999

One time report, not a subscription, no integration required. You send recorded calls, we send back a scored report.

Why this exists

We built our own voice agent and ran this same stress test on it first: people talking over the bot mid sentence, both voices live at once, interruptions it usually caught late, after talking over us for a bit. We scored it with the same harness: it talked over the caller in 3 of 5 barge-in events and took a median of about 7 seconds to actually respond once someone cut in. So we turned the same test into a report you can run on your own agent, scored the same way, on calls you send us.

This is a report on how your agent behaved on the calls you send us. It is not a claim about our own technology, and it is not a comparison against any other product. If your agent already handles interruptions fine, the report says that. We are not claiming to be the only place that tests for this. What we are selling is the format: one time, $499, a scored number back this week, no platform to sign up for.

Who this is built for

Teams running a cascaded voice agent stack, Vapi, Retell, Bland, Pipecat, or self-hosted LiveKit, who want a barge-in number without adding another platform or an ongoing subscription.

If you are already on LiveKit Cloud with its adaptive interruption handling turned on, read the FAQ below before you buy. This test may tell you less than you'd expect, and we'll say so before you pay.

What the stress test covers

Three specific things, nothing broader.

Barge-in

What happens the moment a real caller starts talking while your agent is still talking. Does it stop, or does it keep going.

Overlap

How your agent behaves during the stretch where both voices are live at the same time.

Interruption recovery

After the interruption, does your agent pick up what the caller actually said, or does it ignore it, repeat itself, or answer something else entirely.

What's in the report

Fixed, objective criteria only. Nothing you'd have to take our word for.

timestamped Every barge-in and overlap moment we find across your submitted calls, with a timestamp so you can go listen yourself.
yes / no Did the agent yield when interrupted.
seconds How long it took the agent to respond after the interruption.
yes / no Did the agent talk over the caller.
summary A one page scorecard rolling all of your submitted calls into a single view.
format Delivered as a PDF, reviewed by a person on our end, not a fully automated dump.

No subjective "handled it well" judgment calls anywhere in the report. Every line is something you could verify yourself by listening to the clip.

How the test actually runs

You export 10 to 15 recorded calls from your own stack (Vapi, Retell, Bland, Pipecat, self-hosted LiveKit, or any call recordings you can hand us).
If you need an NDA before sending anything, say so and we'll sign one first.
We go through every call and mark every barge-in and overlap moment we find.
We score each moment against the three fixed criteria above. Human reviewed, model assisted, so it moves fast without being a black box.
You get the report back by email, usually within a few business days depending on how many reports are ahead of yours in the queue.

Data and trust

✓ Recorded calls only. This test does not touch your live system and does not need API access.
✓ You choose which calls to send. Ten to fifteen is enough for a real read.
✓ NDA available before you send anything, just ask.
✓ We ran this on our own agent before offering it to anyone else, so nothing here is a claim we haven't already tried on ourselves first.

Pricing

Founding cohort

$499 one time

Limited to the first 10 companies. Same report as standard.

Standard

$999 one time

After the founding cohort fills. Same report, same process.

No seats, no setup fee, no monthly platform charge either way, this is a one time report, not a subscription.

Mention only, not pushed here: once you've seen your first report, there's an optional $299/month way to re-run the check whenever you ship a change to your agent. We only bring that up after delivery, never before.

Get your report

After you submit the form below, we reply within one business day with a payment link and instructions for sending your call recordings.

Request your report

Tell us where to send the payment link. We reply within one business day with a payment link and upload instructions, we do not charge you anything on this page.

FAQ

What counts as "barge-in"?

Any moment a real caller starts speaking while your agent is still speaking, whether they're asking a question, correcting the agent, or just talking over it.

Do you need API or live access to my agent?

No. This test runs on recorded calls only. You export the audio, we work from that.

Which platforms do you support?

Calls exported from Vapi, Retell, Bland, Pipecat, or a self-hosted LiveKit stack all work fine. On something else, send a sample call first and we'll tell you if it works.

I'm on LiveKit Cloud, does this apply to me?

If you're on LiveKit Cloud with adaptive interruption handling turned on, this test will tell you less, since some of what we score is already handled for you. We will say so before you pay, not after. If you're not sure whether that's on for your setup, send a sample call first and we'll tell you.

Is the scoring automated?

Human reviewed and model assisted. A person on our end goes through every flagged moment before anything ships in your report. It is not a fully automated pipeline and we won't tell you it is.

What if my agent already handles interruptions fine?

Then the report says that. The scoring is objective (yield yes/no, response time, talk over yes/no), so there's nothing to spin either way.

What happens after the first 10 spots are gone?

Price moves to $999 one time. Nothing else about the report changes.

What's the $299/month for?

An optional recheck any time you change your agent. We only offer it after you've seen your first report, never before.

How long does it take?

We review calls by hand, so turnaround depends on the queue. Usually a few business days. You'll get a specific date after checkout once we see your calls.

Find out before your customers do.

Ten to fifteen recorded calls is all it takes to get a real read on barge-in, overlap, and interruption recovery.

Get your report, $499