Skip to main content

​
At a glance

You don’t need the web app to detect tests. A generation session started from the REST API, MCP, a pull request, or Slack detects what to test and writes the tests together (see Generate Tests).
  • Detection runs as part of generation: sessions started from the REST API (POST https://api.checksum.ai/public-api/v1/auto-generate), MCP (checksum_test_generate), a PR comment, or Slack detect and generate together. The web app can also run detection on its own per collection.
  • Source code improves results: detection reads routes and components from the connected code repository. The approach is picked automatically.
  • Output is definitions, not code: each flow has a title, a description of the steps, and a start URL.
Developer guide
Automate coverage for code changes

​
Automating detection

You don’t need the web app to automate detection. The agent detects what to test and generates the tests in the same session when you start generation from the REST API, MCP, a pull request, or Slack. To automate coverage for code changes, use PR-based generation (POST https://api.checksum.ai/public-api/v1/auto-generate, or auto-generate on PR open).

​
Detect flows from your coding agent

With the Checksum MCP server connected, your coding agent can run detection on its own, without generating tests yet:
Your agent calls checksum_detect with your instructions: which areas to explore, what to skip, and how to log in. It can file the specs under a collection and let the agent read your connected code repositories too. It returns a sessionId your agent polls with checksum_session_status. When the session finishes, the flows are ready to review, and your agent can generate tests for exactly those flows with checksum_test_generate (by collection or by userStoryIds). Pass deepMode: true for a Deep run that pauses for plan approval. See Typical flows.
▦
In the Checksum web app
Run detection per collection and review the flows it finds

​
Auto-detect flows for a collection

1

Open Test Generation

Choose Test Generation in the sidebar of app.checksum.ai.
2

Pick a collection

Select an existing collection or create one for the feature area you want covered (e.g. “Checkout Flows”).
3

Click Auto-detect test flows

Choose Standard or Deep (see Deep vs Standard detection). An agent session starts.
4

Answer questions

The agent may pause to ask about your app, e.g. “Which features should I focus on?” or “Are there any areas to exclude?”. Answer them in the session view and it continues.
Auto-detect test flows button on a collection

The Auto-detect test flows action on a collection.

​
Review detected flows

When detection finishes, the flows it found appear in your collection. Each flow has:
  • A title describing the user journey
  • A description of the steps involved
  • The start URL for the flow
A detected flow card

A detected flow, ready to edit or generate.

For each flow you can:
  • Edit the description or steps
  • Delete it if it isn’t relevant
  • Add manual flows for scenarios the agent didn’t detect
  • Generate tests from it (see Generate Tests)

​
Detection vs manual flows

Both produce the same thing: test flow definitions you can generate from. You can combine them. Detect flows automatically, then add a few manual flows for edge cases. To cover a single pull request, you can skip flows entirely and generate from the PR.
i
How it works
What detection looks for and how to improve its results

​
What detection finds

Detection scans your application to discover user flows you can test, like “User can sign up”, “Admin can create a report”, or “Customer can complete checkout”. The agent looks for:
  • Key user journeys through your application
  • Critical business flows
  • Common interaction patterns
  • Edge cases and error scenarios
The output is a set of test flow definitions that describe what to test, not code. Each flow then feeds generation, where the Playwright code is written.
Test Generation page showing the detection menu for a collection

Detection runs per collection from the Test Generation page.

​
Use your source code

Detection reads your actual routes, components, and interactions from your connected source code repository. That makes it much more accurate and faster than exploring the rendered UI alone. Checksum connects your code repository during onboarding, using the read-only access you grant in the onboarding checklist. After that, you can change it at any time in Settings → Git Integration. Checksum picks the detection approach automatically, based on whether a code repo is connected. There’s nothing to configure.

​
Deep vs Standard detection

See Deep vs Standard Modes for how the modes work across every pipeline.

​
Troubleshooting

Start generation instead. The agent detects what to test and generates the tests in the same session. To automate coverage for code changes, use PR-based generation (POST /auto-generate, or auto-generate on PR open).
Check that your code repository is connected (Settings → Git Integration). Re-run in Deep mode and give detailed answers in the interview. Name the features to focus on or exclude in your answers.
Check the environment URL in Environments. If it’s behind a VPN or firewall, confirm Checksum’s IPs are allowlisted (Security & Access).

Generate Tests

Turn flows into Playwright tests.

Agent Sessions

Answer the agent’s questions.

Deep vs Standard

How thorough the analysis is.