Skip to content
Elenchos
Install
QuickstartVerificationKaneMCPAgentsEvidenceSecurity

Kane is the only verifier.

Elenchos consumes Kane structured output. It never replaces Kane with a mock, and it never treats coding-agent narration as proof.

Install and authenticate

Terminal
npm install -g @testmuai/kane-cli
kane-cli login
kane-cli whoami
kane-cli balance

`whoami` and `balance` are readiness checks. Kane credentials stay in Kane's own local session. If the Kane shim is not on PATH, set `KANE_CLI_PATH` to the installed entry file.

Author a Functional test

Elenchos expects a stable Kane test as part of the verification contract. The author command uses Kane's structured generation flow.

Terminal
npx elenchos author path/to/task.json --output tests/feature_test.md
  • Sends the task and acceptance criteria to `kane-cli generate ... --agent`.
  • Preserves Kane's scenarios, cases, request ID, and clarification state.
  • Saves with `generate --save --req ... --agent`.
  • Copies exactly one generated `_test.md` file to the output path.

It stops when Kane saves no Functional test or more than one. That leaves the selection with you.

Terminal
npx elenchos author path/to/task.json --refine "Use the local task list page" --request-id <request-id> --output tests/feature_test.md

Put criterion IDs in the Kane test

tests/add-task_test.md
---
mode: testing
max_steps: 50
timeout: 60
target: chrome
headless: true
---

# Session: add-task

## AC-001 AC-002 Add and display a task
Go to http://127.0.0.1:3000, add a task named "Ship the invoice export", and assert that it is visible in the task list.

Elenchos maps a criterion only when structured Kane events or the matching result step identify that criterion. A criterion that cannot be mapped remains UNVERIFIED.

How Elenchos runs Kane

During run or verify, Elenchos executes `kane-cli testmd run <test> --agent`, parses NDJSON events, classifies the outcome, and stores per-attempt stdout and stderr under `.elenchos/runs`.

  • Failed or incomplete `test_md_done` events take precedence over inner `run_end` events.
  • A timed-out run needs a confirmed product verdict before it can become FAIL.
  • Credit, duration, screenshot, and dashboard fields are kept when Kane reports them.
  • Credential-like strings are scrubbed before JSON is saved or returned.

Larger requirement documents

For a PRD, use Kane's assurance flow, then hand Elenchos the accepted saved Functional test.

Terminal
kane-cli context ingest ./PRD.md --mode agent
kane-cli design tests --use-case <use-case-id> --mode agent
kane-cli testmd run tests/feature_test.md --agent