Summary
The tedious half of test writing, handed to something that does not forget a boundary.
You describe the field ranges once and get back every boundary and its neighbours, every valid and invalid category member, combined pairwise when the parameter count makes that sensible. The assertions are left to you, so it produces a scaffold rather than a finished suite.
What it is
A server with a single tool: give it a JSON specification of fields and value ranges, and it returns Vitest test code built from boundary-value analysis and equivalence partitioning.
What you get
- generate-test-code, which turns a JSON specification into runnable Vitest code
- Boundary values generated around every continuous range — just below, at, just above, and the middle
- Discontinuous ranges handled properly, including the gap values that fall between two valid windows
- Categorical fields expanded into both their valid and their invalid members
- Pairwise combinations once a specification has enough parameters that the full cross-product would be wasteful, and all combinations when it does not
- Output in the test.each form, split into valid and invalid blocks with markers where your subject under test goes
- Prompts explaining how to write the specification, with boundary and equivalence examples
Requirements
Nothing — no account, no key. Runs under Node.
Setup effort
One command — npx testing-mcp-vitest