PROOF. Spec Lab Preview About Proof ↗

Free developer tool · By Proof

Generate test cases for MC/DC coverage

Plan Modified Condition/Decision Coverage (MC/DC) from a written requirement. Review the Boolean formula, generate concrete inputs and expected outputs, and inspect each condition’s independent effect.

Analysis of your requirement model. No application code is executed.

From specification to test table

Make the decision explicit.
Make the cases concrete.

  1. Interpret the rule

    Paste one decision with its conditions, limits and exceptions. AI extracts a reviewable formula and variable meanings. When a policy is incomplete, the tool can ask a clarification question.

  2. Generate the cases

    Native Proof calculates feasible inputs and expected outputs from the interpreted rule. Boolean combinations and numeric predicate regions become concrete test rows.

  3. Inspect, edit and export

    Check the formula against your specification. Edit it to regenerate without another AI call, then copy a formatted table or export Markdown and CSV for your test suite.

The method, in context

What is MC/DC coverage?

MC/DC stands for Modified Condition/Decision Coverage, also written MCDC. It combines condition coverage, decision coverage and evidence that each condition can independently affect the decision’s outcome. For unique-cause independence pairs, change one condition while holding the others fixed.

This MC/DC table generator checks independent condition pairs in your interpreted requirement model, where feasible pairs exist. Code-level MC/DC measures what tests exercised in the implementation. Expand the methodology beside your results to inspect the model’s pairs, formula and input assumptions.

Read NASA’s multi-condition requirements guidance · Requirements-based testing

Is this MC/DC coverage of my code?

These cases come from your requirement model. The tool does not inspect or execute your application. Measuring code-level MC/DC requires instrumented execution against the implemented decisions. See a measured code-coverage example.

Does the table show the minimum number of tests?

The table contains representative cases for the supported Boolean combinations and numeric predicate regions. Its size is not a proved minimum or a complete test count for your feature. Run the cases against your implementation, and review acceptance journeys and failure handling separately.

How does MC/DC differ from decision coverage?

Decision coverage shows that a decision produced both true and false. Condition coverage shows both values of each condition. MC/DC additionally requires evidence that each condition independently affected the decision. A test suite can cover both decision outcomes while missing that independence.

When should I use full analysis?

Quick mode focuses on one stateless Boolean decision with Boolean conditions and numeric limits. It does not perform hazard analysis or develop a broader verification plan. Choose full analysis from the tool for requirements, open decisions, hazards, coverage gaps and evidence to collect. It takes several minutes and starts only after you submit the specification.

Worked MC/DC example

Three conditions. See what each one changes.

A release is allowed when checks pass and either a reviewer approves or an emergency override is present.

allowed = checks AND (approval OR override)

Four selected cases for the example decision. These are expected outcomes, not executed test results.
CaseChecksApprovalOverrideAllowed
1FalseTrueFalseFalse
2TrueTrueFalseTrue
3TrueFalseFalseFalse
4TrueFalseTrueTrue

Independent condition pairs

Checks: cases 1 and 2
Only Checks changes. Allowed changes from False to True.
Approval: cases 2 and 3
Only Approval changes. Allowed changes from True to False.
Override: cases 3 and 4
Only Override changes. Allowed changes from False to True.

A complete truth table has eight rows for these three independent Boolean inputs. These four selected rows demonstrate the three independence pairs. The generator returns representative truth-table cases and the pairs it can establish; it does not promise a minimum test suite.

Generate cases for your own decision

Measure MC/DC coverage in Go

Take the next change into your repository

Your product is more
than one decision.

Keep the intent and the evidence together as it changes.

Proof connects your product’s requirements across components to the code, tests, hazards and evidence behind them. Start with one component, then follow its dependencies and the behavior a change may affect.

Work in your repository with Claude Code, Codex or your existing coding agents, using your project’s requirements and recorded decisions as context. Give the next task the requirements to preserve, the checks it needs and the evidence a reviewer can inspect. Your project policy keeps approvals and exceptions under your control.

Your saved project

Opening your test plan…

Loading your specification, answers and saved results. No new analysis is being started.

Delete project

Delete this project?

Downloaded copies and records already held by AI providers are not deleted by this action. Non-content usage counters are retained separately.

From intent to testable behavior

Know what to test. And why.

Start with a feature specification. Resolve ambiguous rules, see the risks, and leave with a test plan your team can use.

One feature, its rules and its exceptions.0 / 12,000

Add a specification or try the example to begin.

Checking service availability…

Private by default. Not used by Proof for AI training. Saved analysis records stay until you delete them. Source and answers are not sent to analytics.

Your specification, answers and generated review are sent to the configured AI provider. Leave out secrets. No repository access or application test execution is needed.

How your data is handled

Current model configuration: not yet available. Recorded request identities are included in the report.

Your specification, answers, generated requirements, test plan and project files are saved until you delete them. Use “Delete project” to remove a full-analysis project. Delete quick analysis records from Your runs. Draft and shared links stop working when the project is deleted. Download a copy whenever you need one.

We use your content to generate and review your plan. OpenRouter requests are restricted to model providers whose policies disallow training on inputs and outputs. This is separate from provider logging and retention; provider data policies apply.

A signed browser cookie associates Your runs and the one-active-run rule with this browser. This list is private to the browser identity. Your full-analysis draft link reopens the project in any browser until you delete it. Keep the draft link private.

Deleting removes the saved Proof project, not downloaded copies or records already held by the AI providers. Spec Lab sends only usage events, elapsed times and counts to PostHog and Google Analytics; it disables session replay and automatic interaction capture. Source, answers, generated text and private links are not sent to analytics. Non-content usage records, such as token counts and cost totals, are retained separately for billing, limits and abuse prevention. Provider retention is separate from Proof’s project storage. Website privacy · Ask about data handling