Everything Checked tests, and how
APIs, websites, load and real user journeys. One plain-English report at the end of each.
API testing
Bring your API docs. Checked reads them, explains what the API does, proposes a test plan and writes the tests, then runs them against the environment you pick.
- Imports OpenAPI 3 and Swagger 2 (link or file), Postman and Insomnia collections, GraphQL introspection, HAR recordings, cURL commands and plain endpoint lists.
- Generated suites: regression, contract (responses checked against your schemas), edge cases, security (auth stripped or forged, IDOR probes, injection strings), headers and CORS, error handling and availability.
- Deeper suites on paid plans: performance budgets, idempotency, pagination, resilience, rate limiting, fuzzing, caching, decision tables, pairwise, state transitions, end-to-end scenarios and compliance packs that cite the clause each case evidences.
- Auth that works the way your API does: bearer tokens, API keys, basic auth, or a login flow that fetches a token first. When sign-up is open, Checked can register a throwaway test account itself.
- Every result keeps the request, the response and each check, with a diff against your baseline run so you see what changed.
- Write or edit cases by hand, or import them from Excel, and let the AI repair tests that fail for the wrong reason.
Website checks
Give Checked your site's address. It crawls the pages it finds through links and your sitemap and checks each one in a real browser.
- Broken links and redirects, inside your site and out.
- Console errors and failed network requests on every page.
- Accessibility violations, performance timings and SEO basics.
- Security headers, cookies, forms and how pages hold up on small screens.
- Screenshots of each page, with visual changes compared against an earlier run.
- A free teaser scan of any site you own, with no account, at the free site scan page.
Load testing
Find out how your system behaves under real traffic before your users do. Load tests run on k6 and report latency percentiles, throughput and errors as the test runs.
- Smoke, ramp, spike, stress, soak and breakpoint profiles.
- Large runs are split across parallel test agents on the Business and Team plans.
- Big tests only run on domains you have proved you control, with a header, a file or a DNS record. Unverified targets keep a small safety cap.
- Tests stop early by themselves when most requests are refused, so a firewall never gets your address flagged.
- Ready-made allowlist rules for Cloudflare, nginx and AWS WAF, and traffic that always says who it is.
AI explorer and user flows
An AI tester drives a real browser through your app the way a person would, and writes down everything that goes wrong.
- Signs in with the test credentials you give it, fills forms and follows journeys.
- Records every finding with the steps to reproduce it and a screenshot.
- Describe a user flow in plain English (sign up, add to cart, pay) and replay it on every run.
- Or record a flow yourself: start the browser recorder on your site, click through, and your steps are saved.
- When it hits something only you can answer, such as a one-time code, it asks you in real time.
GitHub fix PRs and PR checks
Connect GitHub once and testing becomes part of every change.
AI fix pull requests are included each month on paid plans: 5 on Starter, 40 on Business, unlimited on Team.
- Runs your tests on every push to the branches you choose.
- Posts a commit status and a summary comment on pull requests, with the issues to fix before merging.
- Drafts a fix for the issue you pick and opens a pull request on a new branch for you to review. Nothing is merged for you.
- Fix pull requests are only opened when you turn them on for a project.
Reports and certificates
Every run ends with a report written for the whole team, not only the person who ran it.
- A health score out of 100 and a release answer: safe to ship, ship with care, or hold the release.
- Failures grouped by root cause, explained in plain words, with a reproduction and a suggested fix.
- What was not tested and why, so silence is never mistaken for safety.
- Ask the AI about a run and get answers grounded in its results.
- Share a read-only link, print or save as PDF, export to Excel or Gherkin, and issue a release certificate that records whether your own release gates were met.
- Send findings to Jira or Linear, and get results by email or in Slack.
Which features each plan includes is on the pricing page.
See what Checked finds in your product
Start on the free plan. No card needed.