What is AGEST ZERO TEST?
Point it at the URL of your system and it runs end to end — automatic screen exploration, test design, execution and comparison — detecting the differences (regressions) between the version before your change and the version after it.
There is no test code to write and no specification to prepare. It runs on an on-premises AI that stays entirely inside your own network, so even highly confidential systems can start using it the same day.
Three characteristics
All three follow from running it inside your own network.
No security worries
It uses only the AI you host in-house. Nothing is ever sent to an external AI service. The data under test and the queries to the AI all stay inside your network. It can be used with highly confidential systems.
No token budget to worry about
Because it uses your in-house AI, there is no metered billing to an external service. Run it as many times and for as many hours as you like without watching the cost climb. You never have to abandon a nightly run because of what it would cost.
It runs from a URL alone
No specifications, no test code. If you know the URL of the target, you can start the same day. For systems that require login, please provide a test account.
Three more that only AGEST — a testing specialist — can offer
This tool is not built from test automation technology alone. It carries the accumulated experience of a company that has specialised in verification.
We built our verification experience straight into the tool
AGEST has specialised in software verification for many years. The perspectives built up on real projects — "in this industry, this is what you look at" — are built into the tool for 24 industries. For each industry it holds the checks that matter, the boundary values, the abnormal-case patterns and the business-rule checks.
It identifies the industry from the screen structure of the target system and automatically adds the checks that industry tends to be asked about. Instead of a person having to remember and write them, the tool remembers them.
The effect shows in the numbers. Building these business checks into test generation raised the agreement rate with AI-produced test design from 66.4% to 76.6%. (Measured internally, 16 August 2026.)
Rules and AI, tuned together
Test generation is not left to AI alone. "Rules" assemble steps deterministically from the type and length of each input field; "AI" works out combinations of values that hold up as real business data. Each has its own role.
It also runs with AI switched off. And whatever the AI produces is checked against the elements that actually exist on the screen before it is accepted — so it never invents an operation that isn't there.
Rather than leaning on one or the other, we have tuned how to mix the two while measuring the results. That mix is our own tuning.
AGEST's own technology
There are two things at the core of this tool: how to compare the output before and after a change, and how to walk the screens far enough to reach the paths where real business is completed.
A mechanism that tells apart values that change on every run — timestamps, document numbers and the like — from values that changed because of your fix. And a mechanism that adjusts the order in which screens are walked so that it reaches the core of the business flow. Both were developed in-house and are patent pending in the United States.
How a test runs
Design the test automatically
Enter the target URL and the tool walks the screens by itself, records the operations and builds test cases. The checks that business systems are commonly asked about are added automatically.
Review the contents
You can review the generated test cases on screen. The list can also be exported in a format that opens in a spreadsheet.
Re-run against the changed version
Point it at the URL of the changed version and run the same tests again.
Report the differences
You receive what changed as a list, as diagrams and as a report. The only thing people review is those differences.
What it cannot do
There are four things it cannot do by design. We would rather say so up front.
Judging whether something was correct in the first place
It is a tool for seeing whether anything changed from before. It cannot find something that was already wrong before.
Changes that never reach the screen
If only the data changes and the display does not, it cannot see it. That is what for Data covers.
Recalculating like a calculator
It cannot work out the correct answer to a payroll calculation itself and verify it.
Defects caused by data or configuration
Errors in master configuration or in already-registered data cannot be produced from screen operations.
Scope of support
| Target systems | Business systems operated through a browser (in-house systems, web services). Supported browsers: Chrome / Edge class (Chromium-based). |
|---|---|
| What you provide | The URL of the system under test. A test account if login is required. |
| Where it runs | Installed and run inside your own network. Data from the target system never leaves it. |
| Output | Test case list, comparison dashboard, report, screen transition diagram. |
| Out of scope | Processing that cannot be touched from a browser, such as overnight batch aggregation. |
Want to know whether it works with your own system?
Check the detailed specifications and scope, or talk to us about your environment.