The enterprise apps your teams run on.
Regression before every upgrade and release. Then the same tests on a schedule, watching the journeys people depend on every day.
- ServiceNow
- SAP
- Siemens Teamcenter
- Citrix
- Vendor portals
Agentic testautomation.
Scale to thousands of tests a day with a system that stays stable, safe and self-improves across every device and regulated environment.
Agentic speed. Enterprise safeguards.
# Radio preset survives a source switch
## Preconditions
- Ignition on. Radio is the active source.
## Steps
- Open Radio from the home screen.
- Long-press the tile FM 98.4. Expected: “Saved to preset 2”.
- Switch the source to USB, then back to Radio.
## Postconditions
- Preset 2 plays FM 98.4.
Why AskUI
Scale with agents,
not headcount.
AskUI scales your testing team with agentic test automation written in plain language. It stays stable across every device and legacy codebase, with the audit-ready on-premise capabilities that regulated environments need.
PromptForge · ToolForge · Grid
Test faster, remove bottlenecks
Automate the repetitive clicking and keep the judgment calls: edge cases, exploratory testing, and deciding what is right.
Cortex · Dreaming
Stop babysitting and maintaining
AskUI tests from pixels, not selectors, so nothing breaks when a button moves.
Grid
Every UI, device and screen
From 20-year-old Windows apps to in-car HMI screens, in industries where nobody is going to open the APIs.
ToolForge
Tests in plain language
Bring your existing tests from spreadsheets, CSV, TestRail or Helix ALM. Testers extend the agent themselves.
Airlock
Secure, certified, compliant
On-premise or fully air-gapped, with your own model. We never train on your data. ISO 27001 and GDPR.
Dreaming
Self-improving
Between runs, Dreaming studies every pass and fail and suggests small fixes for you to approve.
Report · watchdog
Trust every result
A step-by-step report with screenshots, the agent's reasoning and a failure analysis. A watchdog stops runs that loop or drift.
Systems self-improve. Tests stay stable. So where does it run?
Where it runs
One agent for the apps you run
and the software you ship.
Enterprise apps your teams depend on, and the software that leaves your factory on a machine. Same agent, same test format, same evidence.
The software you ship.
HMIs, head units, configuration tools and companion apps, tested on the real device before it leaves the factory. Over HDMI and USB, with nothing installed on the target.
- Machine HMIs
- Head units
- Configuration tools
- Companion apps
- POS terminals
If a tester can see it, AskUI can test it. Even where nothing can be installed.
- ServiceNow, SAP, Teamcenter
- Desktop apps
- Head units
- Machine HMIs
- Citrix & RDP
- POS & kiosks
- Canvas UIs
- Android & iOS
One agent, every screen. So what do your testers have to write?
How it works
Plain language in.
Markdown, CSV, even a scanned PDF spec. Kept in your Git repository. The same format for a ServiceNow form and a head unit.
# P1 incident reaches Network Ops
## Steps
- In ServiceNow, create an incident with impact High, urgency High.
- Set the category to Network and submit. Expected: priority shows “1 - Critical”.
- Open the incident from My Groups' Work as a Network Ops agent.
## Postconditions
- Assignment group is Network Ops.
Every step. Proven.
When a test fails, you see why. Without running it again.
1
2
TC-1143 Presets survive an ignition cycle
- 1Save FM 101.7 to preset 3PASSEDObserved Toast reads “Saved to preset 3”
- 2Cycle the ignition off and onPASSEDObserved Home screen shown again
- 3Open preset 3FAILEDExpected Plays FM 101.7Observed Preset 3 is empty
TC-1143_report.md · step_3.png · conversation.json
Proven on every run. But does it get better the more it runs?
It learns
AskUI Research · CUA-Cortex
Most agents forget.
Cortex remembers your app.
Without memory, an agent relearns your app on every run. Cortex keeps what each run learned. So every run after is faster, cheaper and more reliable.
Run 1 · Explore
The first run explores.
No DOM, no labels it can trust. The agent tries Payment, hits a bare “S02” on Refund, backtracks and finds the manager PIN on its own.
- Raw run: screenshots and actions
- Steps each one a model call
- 1,423
- Agent time over the whole evaluation
- 100%
- Repeated mistakes blocked by a rule it had met
- 87
- Tasks passed of 18, every run
- 12/18
Without memory: the agent on its own.
What one run learned, every tester gets. Shorter test cases, generated ones, a coverage estimate, a manual and answers.
A coverage estimate, for free.
From the runs you already have: which screens and controls nobody has operated yet. Computed from the map, without a model call.
Documentation that writes itself.
One user manual per application, kept current by every run. People read it. The agent searches it.
Test cases from real behaviour.
Expected results come from what a run actually observed, not from guesses.
Answers for testers.
Ask how to do something. Every control in the answer is checked against memory. One it doesn't know is dropped, never invented.
Built from screenshots and actions alone. No source code, no DOM, no accessibility tree, so it works on every surface AskUI can see. Figures from the AskUI research evaluation: 12 teaching and 18 testing tasks in three industrial-style test apps, three runs each.
Results
Measured in production.
Published with the customer.
Security
Nothing leaves
your network.
Airlock seals the whole platform inside your network. Your own model, on your own hardware, even air-gapped.
- Your modelClaude Haiku 5.5, Azure or self-hosted
- Offline licenceNever calls home
- Secrets redactedThe model never sees them
- ISO 27001GDPR · Made in Germany
Built for the way
enterprises buy.
Proven on your system before anyone signs. Scaled without seat counts.
- PilotTen of your test cases, run inside your environment, on your systems.
- Roll outYour CI, your model, your network. On-premise or air-gapped.
- ScalePriced per concurrent agent. Unlimited users. Quoted in writing.
Security & deployment Security and data flow (PDF) Architecture (PDF) AI models guide (PDF) All tech specs
Roll out with a partner.Consulting, setup and quality oversight from teams that know AskUI.
Send us ten test cases.
We'll run them on your system.
Or download AskUI Desktop. No sign-up.