Cypress helps make UI tests more dependable by waiting for the state they need, observing or controlling network requests, and encouraging tests that do not depend on one another. It does not eliminate flaky tests: timing races, unstable environments, shared state, and tests aimed at the wrong scope still need to be addressed in the test design.
Why UI tests become flaky
A UI test becomes unreliable when its result depends on timing or conditions outside the behavior it is meant to verify. Cypress identifies animations, API calls, test-server or database availability, dependencies, and network conditions as possible sources of failure. A test may inspect the page before a request finishes, for example, or pass locally but fail when CI runs under different network or resource conditions. Cypress documents these sources of test flake.
The fix is to make the test synchronize on meaningful application state and isolate its assumptions—not to add arbitrary delay until the failure disappears.
Use retry-ability for changing UI, not fixed sleeps
Cypress automatically retries linked queries and their assertions while waiting for the expected UI state. This is useful when rendering or a transition takes time: assert the visible outcome instead of guessing how many milliseconds it will take. Non-query commands, including actions, execute once; Cypress does not repeatedly click or type as though those commands were queries. See Cypress retry-ability.
Quick wins for a faster PC:
Clear out junk files and repair common Windows errorsFree Scan →Scan for outdated or missing drivers - takes under a minuteDriver Scan →Repair Windows errors before they cause bigger problemsFix Now →For example, a test can wait for a button to become visible and assert the result after clicking it:
cy.get('[data-cy="save"]').should('be.visible').click()
cy.get('[data-cy="save-status"]').should('have.text', 'Saved')The assertions express the state the test needs. If the result also depends on a specific network response, wait for that request as described below.
Query retries and test retries are different
Query-and-assertion retry-ability waits for the UI condition within a running test. Configured test retries rerun a failed test, potentially including its hooks. Test retries are disabled by default. They can help expose or contain transient failures, but a test that depends on shared state or an uncontrolled timing assumption remains flawed even if a later attempt passes. Cypress explains test retry configuration.
Synchronize with the request that drives the UI
When content depends on an API response, alias the relevant request with cy.intercept(), wait for that alias, and then assert on the resulting UI. This ties the assertion to the event that matters instead of an arbitrary delay.
cy.intercept('GET', '/api/profile').as('profile')
cy.visit('/profile')
cy.wait('@profile')
cy.get('[data-cy="profile-name"]').should('be.visible')cy.intercept() can inspect request URLs, headers, and bodies; stub a response body, status, or headers; delay a response; and let a test wait for a matching request. Cypress supports combining real and stubbed requests in one test. Use a stub when you need a controlled scenario, such as an error response; use the real service when the integration itself is part of what the test must verify. Avoid intercepting every request with a broad wildcard: Cypress warns that broad interception can add overhead. Details and examples are in the intercept documentation and the network requests guide.
Diagnose failures that appear only in CI
A local pass is not proof that a test is independent of its environment. Cypress identifies network-speed and local-versus-CI differences as possible causes of failures. When a CI test fails, check that it waits for the request that drives the UI, that it asserts meaningful intermediate steps, and that the CI process has not changed application state or reduced access to required resources. Cypress’s debugging guidance describes ways to investigate timing and environment issues.
For recorded CI runs, Cypress Cloud Test Replay is a documented option for examining what happened during execution. Availability depends on the Cypress Cloud setup in use; consult Cypress Cloud’s Test Replay documentation.
Keep tests independent of earlier tests
A test that passes only because a previous test left browser state behind can fail when run alone, skipped, or reordered. Give each test the setup it needs and avoid relying on another test to create its starting state. Cypress documents cleanup of browser context before each end-to-end test, and end-to-end testIsolation is enabled by default. See test isolation and writing and organizing tests.
Do these 3 things before closing this tab:
1Clear out junk files and repair common Windows errors2Fix the driver behind crashes, sound loss and screen glitches3Repair Windows errors before they cause bigger problemsChoose a test scope that proves the right thing
Component, API, and end-to-end tests answer different questions. Pick the narrowest level that can prove the behavior under test, then use other levels where you need integration coverage. Cypress describes these scopes in its testing overview.
| Test type | Scope and feedback | What a passing test establishes | What it does not establish alone |
|---|---|---|---|
| Component | Focused component behavior; Cypress mounts components in a real browser and provides a focused feedback loop. | The component behavior exercised under the test’s setup. | That the complete application’s integrations and journeys work. |
| API | Endpoint behavior without rendering a page. | The endpoint contract or behavior covered by the test. | That the UI renders or the integrated user journey works. |
| End-to-end | An integrated user journey through the application. | The exercised path across the integrated application and its configured dependencies. | That every component, endpoint, or untested path works. |
Component tests offer focused feedback; API tests avoid page rendering; end-to-end tests cover integrated journeys but involve more runtime and exposure to environmental variation. A healthy suite uses levels for complementary coverage rather than treating one passing component test as proof of the whole app. Read Cypress’s component testing guide.
Use accessibility checks as one layer of testing
Automated accessibility scans can flag known rule violations, such as missing labels or poor contrast. Add explicit assertions for the accessible names and semantics your interface intends to provide, and manually assess issues automated rules cannot determine. A scan is not certification or proof of full accessibility. Cypress documents plugin-based checks and Cypress Accessibility, a paid Cypress Cloud offering, in its accessibility testing guide.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.Find the cause of slow suites before optimizing
Measure where time goes before changing tests. Cypress points to the wrong test type, repeated logins, real network calls, bloated CI setup, and resource-constrained machines as possible performance problems. Fix the actual bottleneck: avoid unnecessary network work, limit interception to relevant requests, and remove fixed waits that delay every run without making synchronization reliable. Use retries sparingly rather than turning them into a substitute for diagnosis. Cypress Cloud analytics are documented for slow and flaky tests; see test performance guidance.
ScreenshotNeo as an alternative for screenshot capture
Cypress is for testing application behavior. If you also need a screenshot service for capturing pages outside a Cypress test, ScreenshotNeo is an alternative to try first: it removes known cookie banners, popups, and chat widgets before capture, bills only clean shots, and offers an MCP server for AI agents.
Or skip the browser setup
Make one request to return a screenshot; replace the example URL with the page you need. See the ScreenshotNeo API documentation for options.
curl -G "https://api.screenshotneo.com/v1/shot" -d access_key=YOUR_API_KEY --data-urlencode url=https://stripe.com -o shot.webpCookie banners, newsletter popups, and chat widgets are removed before the shot. Bot checks, blank pages, and failed loads are never billed. Its MCP server lets AI agents take screenshots. The free plan includes 1,000 screenshots a month with no card; paid plans start at $5 for 3,000. Sign up for the free plan.
Frequently Asked Questions
Should I enable Cypress test retries to fix flaky tests?
Retries are disabled by default and rerun a failed test; use them to help handle or investigate transient failures, not as a replacement for correcting the cause.
Can an automated accessibility scan prove my app is accessible?
No. Scans catch known rule violations, but accessibility also needs appropriate assertions and manual assessment.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
