Check a release before it goes live

Pick a build of the simulated agent and run the catalogue against it. Gate sends each customer message to the agent, then asks the system of record what really happened. Each scenario runs three times.

Agent build to test Release candidate 1 A build with defects. Some appear every time and some only in one run of three. Release candidate 2 The same agent after the defects were fixed. The test rules do not change.

System of record Available Gate can read the records and compare them with what the agent said. Not available Gate cannot read the records. See what it does when evidence is missing.

A run takes about ten seconds. One run can be in progress at a time.

Result

Showing run

Gate decision

Download signed evidence (JSON) Open the printable report

Failures by type

Select a type to filter the table below.

By sector

Sector	Scenarios	Failed	Blocking

Scenarios

Select a row to see what the customer asked, what the agent said and what the records show.

Sector Check Severity Status Search

ID

Workflow

Severity

Runs

Status

Check an evidence file

Each run can be exported as a JSON file signed with this service's Ed25519 key. Choose a file you downloaded and Gate will say whether it is intact. Edit one number in the file and try again.

Choose an evidence file

Signing key

The signature shows which service produced a file and that nobody has changed it since. It is not a certificate, and it does not replace the release owner's own decision.

What is real here and what is simulated

Real

The test runner: 100 scenarios, three runs each, 300 executions.
The separate query to the system of record, with its own credential.
The typed checks that work out each result. No scenario carries a pre-set answer.
Reviewer entries, which are added to the record and never change a machine result.
The signed evidence file and its verification.

Simulated

The agent. It is rule-based so that every visitor sees the same result.
The system of record. In a client project this is the client's own staging system.
The policies and amounts. A client's policy owner approves the real ones.
The defects in release candidate 1. They were planted on purpose and say nothing about how often real agents fail.