How ParityDiff handles private APIs and repositories.

Inspect deterministic API analysis, authorized GitHub access, tenant isolation, and opt-in runtime controls. Monitoring and draft delivery do not send source to an LLM.

DATA AND EXECUTION BOUNDARIES

AI proposes. Checks validate. You decide.

ParityDiff approved AI context and independent validation data flowContract and code analysis identify affected usage. AI receives approved source and evidence to propose a repair. Static and configured isolated runtime checks validate it. Your review, CI and authorization control delivery.Authorized inputsContracts + repository selectionAI + static checksApproved model contextScope + type checksConfigured runtimeApproved image + scenarioHardened sandboxChecked AI proposalFindings + proposal checksRuntime evidenceObserved scenario onlyYour review + CISeparate checks, visible limitsAuthorized draft delivery

Authorized inputsAPI contracts and the repositories you select.

AI proposals and static checksAI receives only approved source and evidence.Scope, type and fixture checks validate the proposal.

Separately configured runtimeAn approved image and scenario run in a hardened sandbox.Reports only the observed scenario evidence.

Your review and CISeparate evidence, visible limits, and your delivery authorization.

Authorized draft delivery

AI migration shares approved source and evidence with Meta. Runtime validation requires a configured isolated runner; missing isolation or required evidence blocks it. A finding alone never authorizes a repository write. Your team controls review and merge.