One platform. Author once, run everywhere, trust the result.
Three products on one shared data model: a clean Test Management foundation, an Agentic Author engine that writes and self-heals intent-rich cases, and an Agentic Execution engine that runs real browsers and APIs — then tells you what actually broke.
Test Management your team will actually want to live in
The system of record every agent reads and writes — projects, repositories, suites, runs, evidence and governance. Incumbents are feature-rich but cluttered; here the foundation is clean and low-friction by design, so the team trusts it from day one.
- Structured repository, suites and organized test cases
- Runs, captured evidence and coverage in one place
- RBAC, audit trail and secrets isolation built in
- Integrations and a generic CI/CD hook, not a walled garden
- Agents assist — organize, dedupe and retire stale cases
The Agentic Author engine writes cases — and re-writes them when they break
Author agents generate intent-rich cases from an API spec, a URL, or a user story. Because a case stores its intent — locator-intent plus oracle — the Healer can re-author a broken locator from observed state instead of throwing the case away. Same engine, measured for trust.
- Auto-author API tests from OpenAPI and specs — the uncommoditized edge
- Auto-author responsive-web cases from a URL or user story
- Intent-carrying locators: role, accessible name, nearby text, anchors
- Self-healing = re-authoring, patched with a confidence score, flagged for review
- Healing measured on precision, recall, flake rate and false-heal rate
The Agentic Execution engine runs it for real — and tells you what broke
Executor agents run cases against real browsers and APIs, evaluate the oracles, and capture full evidence. Reporter agents then triage flaky-vs-real, surface a root-cause hypothesis, and compute the regression delta against the prior run — so a red build comes with an explanation, not just a stack trace.
- Real browser and API execution via a pluggable driver layer
- Full evidence: traces, HAR, console logs, screenshots, video
- Flaky-vs-real triage so CI gates on signal, not noise
- Root-cause hypotheses and regression deltas versus prior runs
- Invokable from any CI/CD via a generic hook — CLI, REST, webhook
Built to extend without re-deriving the basics
The MVP is API and responsive Web, depth-first — but the platform is shaped so breadth, engines and governance are already part of the model, not bolted on later.
API + Web now, more on the same model
API and responsive Web ship first, run depth-first. Mobile, desktop and further surfaces layer onto the same orchestration, data model and reporting — sequenced, not promised up front.
Engine- and model-agnostic
A platform abstraction over Playwright, Selenium, Appium and a proprietary API engine, on Claude, GPT or open-source models. Choose the engine and the model — no lock-in.
Enterprise & on-prem
On-prem deployability, audit trails, RBAC and secrets isolation that open-source agent scripts simply don't provide — the Execution plane runs inside your boundary when you need it to.
Bring a spec or a URL. Watch the agents author, heal, and report.
In a short session we'll author cases from your surface, break a locator, and let it self-heal — live, on your data.