Skip to content

Free browser utility

All tools

acceptance criteria generator

Turn known behavior into a reviewable acceptance criteria draft. Supply the feature, actor, action and success result, then add failure behavior and evidence you actually have. Missing answers stay visible.

This utility generates in your browser without uploading input. No signup or AI provider is needed. Use non-secret project context.

Name the behavior or feature to review, not an entire project. Limit: 500 characters.

Who performs this action? Use a role, not a person's private details. Limit: 500 characters.

Describe the known state before the success action. Blank stays unanswered. Limit: 2,000 characters.

Describe one action or event. The same action is used in the supplied negative case. Limit: 2,000 characters.

State the observable expected result. This is a requirement, not a reported test pass. Limit: 2,000 characters.

Supply the negative case's starting context. It replaces the success context for that scenario. Limit: 2,000 characters.

State what should be observable in the negative case. A missing condition or outcome stays unanswered. Limit: 2,000 characters.

Optional reported observations, test results or references. References are not opened and evidence is not verified. Limit: 3,000 characters.

Share warning: your inputs appear in the link, readable by anyone who receives it. Do not include secrets or confidential information. Copying does not send the link or change this page address.

Generated acceptance criteria

Start with the example or enter your own facts. Use Generate criteria when you are ready, then review the result.

Methodology and format

The utility formats your supplied facts deterministically. Feature and actor provide context. The success starting condition becomes Given, the action becomes When, and the expected success result becomes Then. Cucumber's Gherkin reference below explains these roles; write observable outcomes that a reviewer can check.

Supplying either failure field adds a negative scenario using the same actor and action. Its Given uses the failure condition instead of assuming the success context still holds; its Then uses the failure outcome. A missing half is labeled unanswered. If the negative case needs a different actor or action, draft it separately.

Every criterion is labeled untested by this utility. Evidence is quoted as reported and never used to mark a criterion passed. Missing optional fields are listed explicitly. Review and edit the copied or downloaded draft with the requirement owner before implementation or testing.

The Markdown download contains headings and literal fenced text, not an executable Gherkin .feature file. Fences expand to contain user backticks safely. Generation trims surrounding whitespace; internal text remains as supplied. React displays user input as text, without interpreting it as HTML.

Field limits are 500 UTF-16 code units for feature and actor, 2,000 for each behavior field, and 3,000 for evidence. The shared form rejects oversized or non-text values. Share links accept only known fields and a supported format; malformed links are rejected. URLs are limited to 8,000 characters, so larger drafts require file export.

Generation runs in your browser without uploading input or using an AI provider. The utility does not automatically store drafts or log their contents. Copy, export and sharing require your action. A share link contains the inputs in its URL fragment, readable by anyone with the link. Use non-secret context; the website privacy policy covers other website features.

Cucumber: Gherkin reference — Given describes initial context, When the event, and Then the expected observable outcome. Executable scenarios need matching step definitions. Reviewed October 9, 2026.

CommonMark: fenced code blocks — The Markdown export uses fenced blocks to preserve supplied text literally. Reviewed October 9, 2026.

Limitations

This drafts criteria from facts you provide. It cannot discover unstated requirements, judge whether a result is measurable, detect conflicting facts, or establish complete coverage of a feature.

No tests run and no evidence links are opened. A completed form or a supplied success report does not automatically pass any criterion. Check actual results and resolve unanswered fields before treating the draft as agreed acceptance criteria.

One success scenario and at most one supplied negative scenario are included. Other actors, actions, failure modes, security, accessibility and performance requirements need separate review when relevant; none are invented here.

The example is fictional. The export needs editing and matching step definitions before any separate conversion into executable Gherkin. Shared links and exported files are not secure channels for confidential input.

Frequently asked questions

How does this acceptance criteria generator work?

It formats your feature, actor, starting condition, action and expected result into a Given/When/Then draft. Supplied failure details add a negative scenario. It uses fixed formatting in the browser, with no signup or AI provider.

What happens if I do not know a requirement?

Feature, actor, action and success result are required. Missing optional information remains explicitly unanswered. The generator does not invent requirements or infer a failure outcome from the success result.

Does adding evidence make a criterion pass?

No. Evidence is included as a reported statement only. The utility does not run tests or inspect references, and all criteria remain untested by this utility. Record verified results separately after checking the actual behavior.

Can I run the download as a Cucumber test?

No. The download is a Markdown draft with Given/When/Then text for human review, not an executable .feature file. Running a Cucumber scenario requires appropriate Gherkin formatting and matching step definitions in your test project.

Is my input uploaded or included in a share link?

Generation runs in the browser without uploading input. Copy and export happen on request. Copy share link explicitly puts your inputs in a URL fragment that anyone with the link can read; use non-secret drafting context. This describes the utility, not all website requests or scripts.

Related guides and tools

Match completion claims to actual evidence

Review agent work at clear human decision points

Carry agreed criteria into an implementation brief

Define API contracts for AI development

Include criteria and evidence in engineering handoffs

Website privacy policy