Break it in the lab.
Not in production.
Turn real conversations into scenarios the agent must always handle correctly. Run them on every change — and catch regressions before a single user does.
Works on top of the Product Agent.
The Trust Lab dashboard: pass rate, scenario health, and every run across your suites.
Lock in the behavior your users depend on.
Import real conversations
Pull the conversations that matter from production or the Playground. Evals and mock results are auto-generated from what actually happened.
Define what correct looks like
Refine the evals on each turn: what the agent must say, which actions it must call, where it must navigate.
Run on every change
Group scenarios into suites, schedule them daily, and watch the dashboard. A falling pass rate is your regression alarm.
Playground is where you experiment. Conversations show what users did. Trust Lab is where correct behavior gets locked in.
Every failure, explained.
A failed run shows each turn with the agent's actual response next to the eval that judged it — what was expected, and what the agent said instead.
Define correct. Detect drift.
Start from real conversations
Import conversations from production or the Playground — search by user, message, or action. Evals and mock results are auto-generated from what actually happened.
Author multi-turn scenarios
Or build test conversations from scratch: script each turn the user takes, and attach the evals that define a correct response.
Grade what the agent says
LLM-as-Judge for nuanced criteria in plain English, plus Contains, Exact Match, and Regex when you need fast, deterministic checks.
Grade what the agent does
Tool-use evals verify behavior, not just words: the right action called with the right parameters, the right route navigated, the right knowledge retrieved.
Test as any user
Set the simulated user's plan, role, and locale — so plan gating, role-based behavior, and localized responses get tested with real-world context.
Organize into suites
Group scenarios by journey or feature area. Suites are the unit you run, schedule, and track — each with its own pass rate and health.
Run on a schedule
Hourly, daily, weekly, or custom cron — regression detection runs whether anyone remembers to or not, in the timezone you pick.
Drill into every run
Every execution is logged with outcome, pass rate, and trigger. Open a failed run to see the exact turn, the eval that failed, and what the agent actually said. Compare runs side by side to catch flaky behavior.
Never ship the same bug twice
When a real conversation goes wrong, import it immediately. It becomes a regression test that ensures the same mistake can never happen again.
Ready to see Trust Lab in action?
Book a live walkthrough and see real conversations turned into a regression suite that runs while you sleep.
Get a demoWhat teams lock in with Trust Lab.
Ship changes without fear
Change a prompt, an action, or the knowledge base — then rerun every suite and see exactly what it did to quality before users feel it.
Swap models, keep quality
Prove a faster or cheaper model holds the line on your own scenarios — evidence from your product, not a public benchmark.
Catch drift overnight
Knowledge updates, API changes, model updates — a falling pass rate is your signal that something the agent used to get right has broken.
Trust Lab runs the tests. You ship knowing what still works — and what just broke.
Common questions about Trust Lab
How do I join the Trust Lab early access?
Trust Lab is in early access. It runs on top of the Product Agent and tests the agent you've already built: its answers, its actions and its navigation routes. Ask to join the early access, then start with a small suite of the conversations your agent must never get wrong.
How do you regression-test an AI agent?
Turn real conversations into tests. In Trust Lab (early access), you import a conversation from production or the Playground, and it becomes a scenario with evals and mock results generated from what actually happened. Tighten the evals turn by turn, group scenarios into suites, and run them hourly, daily, weekly or on a custom cron. A falling pass rate is your regression alarm.
What kinds of evals does Trust Lab support?
Trust Lab, in early access, grades two things. Response evals grade what the agent says: LLM-as-Judge against success criteria you write in plain English, plus Contains, Exact Match and Regex for fixed facts like prices or legal text. Tool use evals grade what it does: Action Called checks the action and the parameters it passed, Navigate checks the route, and Knowledge Search checks the document it found.
Can I test the agent as different users?
Yes. In the Trust Lab early access, each scenario sets the simulated user before the first turn: user ID, name, email, role, plan, locale and any custom fields your product passes to the agent. So plan gating, role-based behavior and localized answers get tested the way real users meet them, not as one generic test user.
How do I switch AI models without breaking quality?
Treat the model change like any other change: run your suites on it and compare the pass rate with the last run. Trust Lab, in early access, grades both what the agent says and what it does, so you see on your own scenarios, not a public benchmark, whether the new model holds the line. The same suites catch drift after knowledge updates and API changes.
What happens when a scenario fails?
You see exactly where. Runs logs every execution with its outcome, pass rate and trigger, manual or scheduled. Open a failed run and each turn shows the agent's actual response next to the eval that judged it, with the failed assertion highlighted. Fix the knowledge, action or navigation route behind it and rerun the suite. Trust Lab is in early access.
Regressions don't announce themselves. Trust Lab does.
Define what the agent must always get right — and know within a day when anything breaks it.