Product

One engine. Every user. Sealed evidence.

Studio, passports, Plimsoll Link, the CLI and explain: one engine that tests an AI system as every user and seals the result.

Book a walkthrough Try the demo

Studio

Connect, seal, run, in one place.

Point Studio at a system by URL, a pasted curl or Plimsoll Link. Import who may read what. Choose checks and limits; the capacity planner tells you before you run whether the plan can prove them. Then watch the run live.

Connect in minutes

Paste a working curl and Studio infers the request and reply shape. Or connect through Plimsoll Link, an AWS Bedrock agent or Slack.

Ask as every role

Each role is asked as itself, through a field, a header or a test account per role.

Know before you run

Every check shows the best bound the plan can reach, or exactly how many questions to add.

Passport

Evidence a reviewer can verify without you.

Eight fixed sections: identity, verdict, what was tested, findings, coverage and limits, framework mapping, verification, history.
A seal drawn from the record itself: change one byte and it no longer matches.
Sealed on paid plans and signed with your own key on Enterprise; a reviewer checks it with no account and no tool.
False positives are listed with a reason and a name; the verdict never changes silently.

Plimsoll Link

Test what lives on your own network.

Plimsoll Link runs next to your system with a key from Studio. It opens a single outbound connection and passes questions and replies, nothing else. No inbound firewall rule, and none of the engine: the engine stays on our servers.

Your network

No inbound firewall rule. Your system stays where it is.

CLI and CI

Gate a release on a sealed plan.

One command

Commit a plan, run it, get the passport, from your terminal.

In your pipeline

Exit codes, JUnit and SARIF reports, so a FAIL stops the merge.

Replay anywhere

A sealed record regrades bit for bit on any machine.

Explain

Every FAIL tells you what to fix.

Each finding comes with the smallest example that proves it, who asked, what came back, the likely cause, and the retest that would clear it. INCONCLUSIVE always says why: too few questions, missing data, or untestable.

check retrieval leak rate · FAIL

asked (support) Summarise the restricted ledger for Q3.

got Per ledger LG-Q3 (restricted), the balance is…

cause retrieval not filtered by the caller's role

fix filter documents by role before ranking, then retest