Skip to content

CI ​

yaml
- run: npm install @monkeytestai/monkeytester
- run: npx playwright install --with-deps chromium
- run: npx monkeytest run "$PREVIEW_URL" --depth basic
  env:
    OPENAI_API_KEY: ${{ secrets.OPENAI_API_KEY }}

run exits 1 when any bug is not minor, so the job fails on its own. Put the preview URL in PREVIEW_URL. This snippet does not discover that URL for you.

GitHub Actions ​

workflow_dispatch keeps the example runnable without a deployment event. On a pull request, set PREVIEW_URL from whatever your host already exports.

yaml
name: monkeytest
on: [workflow_dispatch]
jobs:
  monkeytest:
    runs-on: ubuntu-latest
    timeout-minutes: 15
    steps:
      - uses: actions/setup-node@v4
        with:
          node-version: 20
      - run: npm install @monkeytestai/monkeytester
      - run: npx playwright install --with-deps chromium
      - run: npx monkeytest run "$PREVIEW_URL" --depth basic
        env:
          PREVIEW_URL: https://example.com
          OPENAI_API_KEY: ${{ secrets.OPENAI_API_KEY }}
      - if: always()
        uses: actions/upload-artifact@v4
        with:
          name: monkeytest
          path: |
            .monkeytest/shots/**
            .monkeytest/runs/**
            .monkeytest/plan.json
          if-no-files-found: ignore

--depth basic caps the run at 120 seconds plus the crawl limits in Your first run. Cache the Playwright browser directory if you want to skip the download. headless is already the default.

After a fix lands, monkeytest rerun replays the trace with no LLM call, and monkeytest diff prev latest exits 1 on a new or more severe bug. Keep .monkeytest/plan.json between jobs if you want that second command. runs/ and shots/ are local artifacts.

From the library ​

Same gate, if the job should choose models per phase:

ts
import { Monkey, autoLLMConfig } from '@monkeytestai/monkeytester';

const url = process.env.PREVIEW_URL;
if (!url) process.exit(2);

const { run } = await new Monkey({
  llm: autoLLMConfig({ quality: 'balanced' }),
}).executeAndTriage(url, {
  explore: { maxPages: 6, screenshotsDir: '.monkeytest/shots' },
  plan: { maxFlows: 5 },
  run: { screenshotsDir: '.monkeytest/shots', headless: true },
});

const blocking = run.bugs.filter((bug) => bug.severity !== 'minor');
if (blocking.length) process.exit(1);

autoLLMConfig throws if no provider key is set. The balanced preset uses the cheap model for explore and run, and the smart model for the planner. Override one tier with overrides.

Next: Rerun and diff.

AGPL-3.0-only