1
Probe
We ask your bot the questions a real customer would, plus the tricky ones.
The pipeline
How Botclipse works
A simple, automated pipeline that tests your chatbot, remembers what "good" looks like, and watches for it quietly going wrong.
Five stages
Probe, gate, judge, baseline, drift.
Every scan runs the same five stages in the same order. Nothing is improvised, which is what makes one month's score comparable to the next.
2
Gate
Safety checks decide what's allowed to run. Nothing risky happens without your permission.
3
Judge
Each answer is scored with exact checks, plus an AI judge for the nuanced cases.
4
Baseline
We freeze a "known-good" snapshot of how your bot should behave.
5
Drift
We re-test on a schedule and alert you the moment it gets worse.
Coverage
The six ways we check your bot can fail
Every probe we run is graded against one of these six categories, so a score you get in July means the same thing as a score you got in March.
Accuracy
Does it give correct facts about prices, hours and policies?
Hallucination
Does it make up things that aren't true or offered?
Boundary violations
Does it promise things it isn't allowed to?
Action integrity
Does it avoid taking actions it shouldn't?
Handoff / escalation
Does it route to a human when it should?
Data leakage
Does it keep internal and other-customer info private?
Why you can trust the results
An audit is only worth as much as its method. Here is what ours is built on.
- Independent: an outside check, not the bot grading itself.
- Fail-closed safety: risky tests are blocked unless you authorize them.
- Real evidence: every finding comes with the actual quote from your bot.
- Continuous: a one-time test can't catch a bot that breaks next month. We do.
Where to next
Get started
Get a free reliability scorecard for your chatbot
Send us the URL and a sentence about what your bot is meant to do. We'll run an independent scan and send back a scorecard across all six failure categories, including the transcripts of anything we break.
Takes about two minutes. Open the request form
No integration, no credentials, no commitment. We test from the outside, like any customer would.