Skip to main content
Powered by Checksum’s world model — built from real user behavior patterns — the agent achieves ~97% test accuracy. Tests include full data setup and cleanup, grounded selectors, and production-ready architecture. Unlike local coding agents, Checksum runs in the cloud 24/7, generating and verifying tests autonomously at scale.
For large-scale test generation, we recommend running detection first to identify all important test flows, then generating tests from the detected flows. This gives you a chance to review and prioritize before generation begins. For quick, targeted test creation, you can skip detection and generate directly from a manually created flow.

​
Overview

When you trigger generation for a test flow, Checksum:
  1. Plans what to build (in Deep mode)
  2. Implements the test — writing both a story file (.checksum.md) and a test file (.checksum.spec.ts)
  3. Reviews its own work for quality and correctness
  4. Verifies the tests pass by actually running them
  5. Opens a PR to your repository with the generated tests
The entire process is autonomous — you trigger it, and come back to a PR ready for review.
Test Generation Flow

​
Triggering Generation

  1. Navigate to Test Generation in the sidebar
  2. Select a collection and find the test flow you want to generate
  3. Click Generate on the test flow
  4. Choose Standard or Deep mode (see Deep vs Standard)
  5. The agent session starts
Test Generation
You can also trigger generation for multiple flows at once within a collection.

​
Trigger from GitHub pull requests

If your code repository is connected via the GitHub App, you can start generation without opening the dashboard.

​
Slash command

Comment on a pull request: 
/checksum generate
Optional instructions on the same line are forwarded to the agent: 
/checksum generate focus on the new SSO login flow and error states
Requirements:
  • The Checksum GitHub App is installed on the repository
  • The commenter has write access to the repo
  • The PR is not from a bot account (Dependabot, Renovate, etc.)
Feedback: Checksum reacts with 👀 while starting, 🚀 on success, or 👎 if the command is rejected. Progress updates appear as a sticky comment on the PR (not as extra reply spam).

​
Auto-generate on PR open (optional)

Your project can enable automatic generation when a non-draft PR is opened (opt-in — ask Checksum support to enable it for your app). The agent runs against the PR diff the same way as /checksum generate, without a manual comment. Draft PRs and bot-opened PRs are skipped.

​
Public API

Trigger generation programmatically: 
POST /public-api/v1/auto-generate

curl -X POST https://api.checksum.ai/public-api/v1/auto-generate \
  -H "Authorization: Bearer <YOUR_API_KEY>" \
  -H "Content-Type: application/json" \
  -d '{
    "prNumber": 42,
    "repoName": "acme-co/webapp",
    "branch": "feature/checkout",
    "metadata": { "source": "release-bot" }
  }'
Poll batch progress with GET /public-api/v1/auto-generate/batch/:batchId. See the API Reference.
Generation from a PR always diffs against the PR head branch — Checksum fetches the head ref from GitHub so feature-branch PRs target the correct code.

​
Generation Statuses

As the agent works, your test flow’s status updates:
StatusMeaning
IdleNo generation in progress for this test flow
QueuedGeneration is queued and will start shortly
RunningThe AI agent is actively generating tests
VerifyingTests are being verified (actually executed)
CompletedTests are generated and PR is ready
FailedGeneration encountered an error
AbortedGeneration was cancelled
EditedGenerated tests were manually edited
PR MergedYou merged the generated tests PR
PR ClosedYou closed the PR without merging

​
What Gets Generated

For each test flow, Checksum generates:
  • Story file (.checksum.md) — A human-readable description of the test with steps, data setup, and verifications
  • Test file (.checksum.spec.ts) — Production-ready Playwright test code
See Story and Test Format for detailed format documentation.

​
Reviewing Generated Tests

When generation completes, Checksum opens a PR to your test repository. You can:
  1. Review the PR on GitHub/GitLab — read the test code, check the story file
  2. Run the tests locally — checkout the branch and run npx checksumai test
  3. Request changes — if something needs adjusting, you can edit directly or re-trigger generation
  4. Merge — when you’re satisfied, merge the PR to add the tests to your suite

​
Next Steps