Core concepts
Lashtest Lab has a small number of concepts that fit together in a fixed hierarchy. Understanding this shape makes everything else in the product predictable.
The hierarchy
Organisation ← your company or team; owns the plan and billing
└── Project ← one body of testing work
├── Suites (nested) ← folders for organising cases
│ └── Test cases
├── Test runs ← cases executed together, with results
├── Test plans ← groups of runs reported on together
├── Run templates ← saved sets of cases for starting runs quickly
├── API keys ← for CI and automation
└── Webhooks ← outbound notifications
In one paragraph
An organisation holds everything and owns the plan. Inside it, a project is one body of testing work — one product or service. Within a project you write test cases and organise them into a tree of suites. To execute testing you create a test run, which is a snapshot of selected cases that you work through and mark with results. If you want to report on several runs as one unit of work, you group them into a test plan.
Concepts at a glance
| Concept | What it is | Notes |
|---|---|---|
| Organisation | The top-level tenant | Owns members, roles, plan, and integrations |
| Project | A container for one body of testing | Has a unique slug; scopes runs, plans, and API keys |
| Suite | A folder for test cases | Nests without limit; cases can also sit outside any suite |
| Test case | A single test to perform | Steps or Gherkin, with priority, status, and tags |
| Test run | A set of cases executed together | The unit you mark pass/fail |
| Test plan | A group of runs | For reporting across several runs |
| Reports | Trends and flakiness | Computed across runs |
Two things that surprise people
A test case is not a copy inside a run. When you add a case to a run, the run references the case and holds your result for it. Editing the case later changes the case — the result you recorded stays attached to the run. This is what lets flakiness and trends be computed: the same case, measured across many runs.
A completed run is immutable. Once you close a run, its results are frozen. That is what makes it a trustworthy record of what happened. If you need to test again, create a new run — and if you want it to contain the same cases, use a run template.
Where to go next
- Organisations and roles — who can do what.
- Projects and suites — organising your work.
- Test cases — writing them.
- Test runs — executing them.
- Pass rate — how the numbers are calculated.