Generating Test Cases with AI

kane-cli generate turns a plain-language description of what you want to test into structured test scenarios and test cases — without writing them by hand and without launching a browser. This page covers AI test-case generation from both the command line and the interactive kane-cli TUI.

A generation produces:

• Test Scenarios — logical groupings of related checks (e.g. "Login", "Checkout").
• Test Cases — the individual checks inside each scenario, each typed Positive, Negative, or Edge.

You generate, review the result, refine it in plain language as many times as you like, and optionally save the functional cases as runnable _test.md files that kane-cli testmd can execute and replay.

note

For the full picture of AI test-case generation — including richer inputs (files, PDFs, issue links) and pushing cases into Test Manager or automation — see the AI test-case generation guide. The CLI here takes a text description and writes local _test.md files. See Limits and scope.

Quick start

# 1. Generate from a description
kane-cli generate "checkout flow on a shopping site"

# 2. Refine it (repeat as needed)
kane-cli generate "also cover an expired card and an out-of-stock item" --refine --req 23271

# 3. Save the functional cases as runnable tests
kane-cli generate --save --req 23271
#    → .testmuai/tests/<suite>/<scenario>/<case>_test.md

# 4. Run them
kane-cli testmd run .testmuai/tests/checkout-23271/login/valid-credentials_test.md

The request id (23271 above) is printed at the end of each generation and is how you continue a request across commands.

The three modes

generate runs one turn per command, then exits — you continue a request by running the command again with --req <id>. (That's the scripted/headless surface; run interactively, kane-cli opens a live TUI session instead — see Interactive mode (TUI).)

Mode	Command	What it does
New	kane-cli generate "<what to test>"	Starts a fresh request and generates scenarios + cases. Prints a request id.
Refine	kane-cli generate "<change>" --refine --req <id>	Adjusts an existing request in plain language — add coverage, narrow scope, change focus.
Save	kane-cli generate --save --req <id> [--out <dir>]	Writes the request's functional cases to _test.md files. No new generation.

--refine and --save both require --req. --refine needs a change description; --save takes no description.

Options

Option	Purpose
--req <id>	The request id to refine or save.
--out <dir>	Where --save writes. Defaults to <cwd>/.testmuai/tests (the same location the TUI uses).
--name <name>	Names the run and the saved suite folder.
--scenario-limit <n>	Maximum number of scenarios to generate.
--per-scenario-limit <n>	Maximum test cases per scenario.
--memory	Use the memory layer — reuse relevant existing test cases and reduce duplicates.
--project <id> / --folder <id>	Test Manager project / folder.
--agent	Emit structured NDJSON on stdout (auto-on when run non-interactively / piped).
--env, --username, --access-key	Environment and authentication — same as kane-cli run. See Authentication.

Interactive mode (TUI)

The commands above are the scripted surface. The kane-cli TUI launches in Run mode (for running tests) by default — switch to Generate mode with /generate, and back with /run. Or jump straight in: running kane-cli generate "<objective>" in a terminal (without --agent) opens the TUI directly in Generate mode and submits your objective.

Unlike the one-turn command-line surface, the TUI is a live session — generated scenarios stay pinned in a Scenarios box that you refine and browse in place:

Input	Action
(type any text)	Refine — describe a change in plain language; the set updates in place
/view [S<n>]	Open the scenario browser — drill scenarios → cases → case detail
/save	Save the functional cases to <cwd>/.testmuai/tests/…
/cancel	Cancel the current generation
/run	Switch back to Run mode

In the scenario browser: ↑↓ move · ↵ open · ◂ ▸ previous/next sibling · x remove a case · ←/esc back. Non-functional cases (Security, Performance, …) show a gray ✓ with a "won't be saved" note — /save writes only functional cases (see Saving is functional-only).

If generation asks a clarifying question, just type your answer to continue. After /save, the files are ordinary _test.md tests under .testmuai/tests/ — run them with kane-cli testmd run.

How a result is shaped

Each generated case carries:

• a title and steps,
• a type — Positive (expected to pass), Negative (tests failure handling), or Edge (corner cases),
• a category — e.g. Functional, Security, Performance,
• a priority.

Scenarios are ordered by importance. The full result is returned at the end of the turn so it can be reviewed before you refine or save.

Saving is functional-only

--save writes only test cases whose category is Functional — those are the ones that translate into runnable _test.md tests. Non-functional cases (Security, Performance, and so on) are still generated and shown, but they are not written to disk. If a request has no functional cases, --save writes nothing and tells you so.

Saved files are ordinary _test.md tests in the standard format — identical to a hand-written one. From there everything in the testmd docs applies: run them, edit them, replay them from cache, commit them to git.

This is the intended path: generate authors test cases → testmd runs them.

Limits and scope

• Input is a text description. File, PDF, audio, and issue-link inputs (and the web product's "Create / Create and Automate") are not part of the CLI.
• Refinement is whole-request. You refine the request as a whole in plain language; targeting an individual scenario or case from the CLI is not yet supported.
• Scenario and per-scenario counts are bounded by --scenario-limit / --per-scenario-limit.

Next steps

• The generate workflow — the new → refine → save → run loop, worked examples, and exit codes.
• Running tests with testmd — run and replay the _test.md files --save produces.
• Authentication — logging in and choosing an environment.