Know exactly what to test, grounded in your codebase
TestHub analyzes your application's repository and proposes reviewable test scenario candidates. Your team decides what becomes a test, and nothing is authored without confirmation.
Coverage should not depend on what someone remembers to ask for
Scenario sets grow organically from individual requests. Discovery replaces intuition with a systematic, repository-grounded answer to the question: what test scenarios should exist for this application?
Before
A tester or developer must know what to test, prompt an agent for it, and review whatever comes back, free-form.
After
TestHub proposes grounded candidates with evidence and confidence. Your team reviews a backlog instead of inventing one.
From repository to reviewed scenarios in five steps
- 01
Analyze the application
TestHub builds a structured understanding of the connected repository: routes, handlers, components, models, validation rules, and existing tests.
- 02
Propose scenario candidates
Opportunities become lightweight scenario outlines: target, preconditions, actions, and expected outcomes, plus rationale, priority, confidence, and repository evidence paths. No scenario code is generated yet.
- 03
Review in the portal or with your agent
Accept, reject, or defer candidates individually or in bulk. Nothing is generated or written without your decision.
- 04
Generate only what you accepted
Accepted candidates, and only those, are drafted into full vocabulary-conformant scenario JSON, validated, and checked again for duplicates before you see them.
- 05
Author into your scenario library
Confirmed drafts are appended atomically to the application's scenario file, and the portal refreshes live. Candidates retain a durable link to the scenarios they became.
Proposal is cheap. Generation is deliberate. Authoring is gated.
Human in the loop
Discovery only proposes. No scenario enters your suite without an explicit accept decision and a final authoring confirmation.
Grounded in evidence
Every candidate carries repository-relative evidence paths. Ungrounded guesses are marked low confidence or discarded entirely.
Cost follows intent
Outlines are cheap to propose and review. Full scenario generation runs only for candidates you accept, never for the whole backlog.
No duplicates
Candidates are deduplicated against your existing scenario file and the active backlog, at outline level and again at draft level.
Runner-compatible by construction
Every candidate is typed API, UI, or mixed and validated against the canonical scenario vocabulary before it can be authored.
Confidence you can see
Extracted facts earn high confidence; convention-based inference earns less. The label is visible before you spend a generation.
Every candidate has a clear, auditable state
Explore all TestHub capabilities · Currently viewing: Agentic Scenario Discovery
See discovery run on your own repository
Request a technical walkthrough and we will run scenario discovery against an application you know well.
Part of SolidX