Vailra is a browser extension that protects sensitive data in your prompts to AI services (ChatGPT, Claude). It detects sensitive items (personal information, API keys, source-code identifiers, financial data) in the text you send, replaces them with synthetic placeholders before the request leaves your browser, and restores the original text in the AI's reply — visible only on your screen.
To do its job, Vailra reads and modifies the outgoing request on ChatGPT/Claude pages and reads the AI's response as it renders. This includes prompt text, detected sensitive entities, the synthetic placeholders, and a local placeholder-to-original map. All of this stays on your device. The map is encrypted in memory (AES-256-GCM) with an ephemeral key that is never saved and is discarded when your browser closes (or within a few hours).
None. The personal edition performs no network requests of its own, sets no cookies, and contains no analytics or telemetry.
storage — save your on/off setting and local counters (on your device).scripting, activeTab + access to chatgpt.com and claude.ai only — inject the on-page protection logic on those sites.sidePanel — show the dashboard.offscreen — run on-device language processing (also fully local).contextMenus — the right-click "toggle protection" entry.A separate enterprise edition offers optional "team analytics" that is off by default. If an administrator explicitly enables it and configures their own endpoint, Vailra sends anonymized metadata only — counts of items masked by category, latency buckets, and coarse timestamps — to that self-hosted endpoint. It never sends prompt content, redacted originals, or the placeholder map. This policy covers the personal edition, which has this disabled and absent.
All data lives only on your device. Uninstalling the extension removes it. The in-memory encryption key is never persisted and expires automatically.
Material changes will be posted here with an updated date.