Guard your prompts
before they leave.
PromptGnome detects 79+ types of sensitive information in your AI chatbot messages and warns you before it's sent. Free to install with powerful Pro features — 100% local detection, no data collection, no trust required.
Watch PromptGnome in action
Type a message containing sensitive info. PromptGnome catches it before it leaves your browser.
Your data stays yours. The AI never sees it.
PromptGnome replaces sensitive data with placeholders, gets the AI's response, then restores the originals for you.
Three steps. Zero data leaves.
PromptGnome intercepts your message locally, scans for sensitive data, and puts you in control before anything is sent.
Intercept
When you hit send on any supported AI chatbot, PromptGnome catches the message before the network request fires.
Detect
A local detection engine scans for emails, SSNs, credit cards, API keys, and 79+ other sensitive data types in under 10 milliseconds.
Decide
You see exactly what was found and choose: edit your message, send anyway, or auto-anonymize with one click. Always your call.
79+ categories of sensitive data
From obvious things like emails and SSNs to subtle leaks like API keys, database credentials, and international government IDs buried in code snippets.
Contact Information
Email addresses, US & international phone numbers, street addresses, ZIP codes, and dates of birth.
Identity Documents
SSNs with area number validation, US passport numbers, driver's licenses, Medicare IDs, and EINs.
Financial Data
Credit card numbers with Luhn validation, IBANs with checksum verification, SWIFT/BIC codes, routing numbers, and crypto wallet addresses.
API Keys & Secrets
50+ token patterns: AWS, GitHub, Stripe, OpenAI, Anthropic, Slack, Discord, Twilio, SendGrid, JWTs, private keys, and high-entropy secrets.
International IDs
20+ country-specific formats: UK NIN, German Tax ID, French SSN, Spanish DNI, Indian Aadhaar & PAN, Canadian SIN, Brazilian CPF, and more.
Network & Infrastructure
IPv4/IPv6 addresses, MAC addresses, database connection strings (MongoDB, PostgreSQL, Redis), and embedded basic auth credentials.
Contextual Detection
Smart trigger-phrase detection for passwords, PINs, passphrases, usernames, and secrets. Also catches env-file patterns and high-entropy variables.
Code-Block Filtering
Smart enough to ignore PII-like patterns inside code fences, URLs, and example data. Dramatically reduces false positives for developers.
Names, Orgs & More PRO
AI-powered detection catches person names, organizations, locations, and medical terms that pattern matching alone can't identify.
Works where you chat with AI
PromptGnome integrates directly with the most popular AI chatbot interfaces.
Powerful protection, transparent pricing
Start free with robust PII detection. Upgrade to Pro for AI-powered analysis, auto-anonymization, and advanced detection modes.
Free vs Pro — side by side
Both plans protect your privacy. Pro unlocks AI-powered detection and auto-anonymization for complete coverage.
| Feature | Free | Pro |
|---|---|---|
| Detection | ||
| Pattern-based PII detection (65+ types) | ||
| AI-powered name & entity detection (+5 types) | — | |
| API keys & secrets (50+ service tokens) | ||
| International IDs (20+ countries) | ||
| Code-block & URL filtering | ||
| Detection accuracy modes (Speed / Balanced / Maximum) | Speed only | All 3 modes |
| Anonymization | ||
| Warning overlay (edit / send anyway) | ||
| Auto-anonymize (replace PII with placeholders) | — | |
| AES-256-GCM encrypted mapping store | — | |
| Response re-hydration (restore originals in AI replies) | — | |
| Platform & Controls | ||
| 8 AI providers supported | ||
| Behavior modes (Warn / Block / Silent) | ||
| Audit log & detection statistics | ||
| File scanning (PDFs, images, documents) | — | |
| Adjustable confidence threshold | ||
| Privacy & Security | ||
| 100% local processing (Speed & Balanced) | ||
| Secure backend NER (Maximum mode — in-memory, zero storage) | — | |
| Works offline (after initial setup) | ||
| Zero data retention (no logs, no database) | ||
| Priority support | — | |
| Price | ||
| Monthly cost | $0 forever | Coming soon |
Pro detection catches what patterns can't
Free uses fast pattern matching for structured data. Pro adds AI-powered NER analysis that understands context, names, and meaning — dramatically reducing missed detections.
Accuracy metrics based on internal benchmarks using the GLiNER NER model evaluated against a synthetic PII test corpus of 2,000+ annotated samples across all supported entity types.
Your data, processed securely — never stored
Pro's AI analysis requires more compute than a browser can provide. Here's exactly how we handle it — with zero data retention, guaranteed.
Zero Data Retention
Your text exists in server memory only during analysis — typically under 400ms. No logs, no database writes, no disk storage. When processing finishes, the memory is immediately deallocated. We couldn't retrieve your data even if asked.
Why Not 100% Local?
Advanced NER models require significant compute. Browsers limit extensions to ~250MB of memory. Pro's Maximum mode uses enterprise-grade models that need dedicated GPU inference — something a browser tab simply can't provide. Speed and Balanced modes still run entirely locally.
You're Always in Control
Backend analysis only activates in Maximum mode — and only with your explicit consent. Switch to Speed or Balanced mode anytime for fully local processing. No data ever leaves your browser without your permission.
Voluntary Telemetry Only
The only data we ever store is anonymised, aggregated telemetry — and only if you explicitly opt in. This helps us improve detection accuracy. It never contains your text, PII values, or anything personally identifiable.
Need custom deployment or team-wide rollout?
We offer custom onboarding, dedicated support, volume licensing, and tailored detection configurations for organizations handling sensitive data at scale.
contact@promptgnome.comPrivacy tools should actually protect your privacy
Most alternatives collect your data, run on remote servers, or charge you for basic protection. We don't.
Generous Free Tier — No Bait-and-Switch
65+ PII types detected for free, forever. Core protection — emails, SSNs, credit cards, API keys, international IDs — never goes behind a paywall. Pro adds AI-powered detection and anonymization for those who need more.
Zero Data Stored — Anywhere
Free and Pro Speed/Balanced modes process everything locally. Pro Maximum mode uses secure backend analysis — but your text only exists in volatile memory during processing (under 400ms), with zero disk writes, zero logs, and zero database storage. We couldn't retrieve your data even if asked.
Sub-10ms Detection, Not Seconds
Cloud-based tools add 500ms–2s of latency per message. Our local detection engine runs in under 10ms — so fast it's invisible. You never feel it working.
Local by Default, Secure by Design
Free and Pro Speed/Balanced modes run entirely in your browser. Pro Maximum mode uses secure backend processing with zero data retention — your text exists only in volatile memory during analysis and is purged immediately after.
AES-256 Encrypted Anonymization
When you auto-anonymize, the mapping between placeholders and real values is encrypted with AES-256-GCM. Session-scoped keys auto-expire. Other tools store mappings in plain text — or on their servers.
8 AI Providers, One Extension
ChatGPT, Claude, Gemini, DeepSeek, Perplexity, Grok, Copilot, and Meta AI — all supported out of the box. Other tools only cover one or two platforms.
Questions? We've got answers.
Pro Maximum mode: For the highest accuracy, text is sent to our secure backend for advanced NER analysis. It's processed entirely in volatile memory (RAM) — no disk writes, no database, no logs. Data is purged the instant processing completes (typically under 400ms). The connection uses TLS 1.3 encryption, and Maximum mode only activates with your explicit consent. You can switch back to fully local processing anytime.
The only other network calls are for subscription verification and optional anonymised telemetry (opt-in, never contains PII).
Pro ($2.99/mo) adds AI-powered NER detection (0.96 F1 score) for names, organizations, locations, and medical terms — entity types that regex physically cannot detect. It also includes auto-anonymization with AES-256 encryption, response re-hydration, 3 detection accuracy modes, file scanning, and 6.9x fewer false negatives. During our launch promotion, all Pro features are available free.
Ready to protect your prompts?
Install PromptGnome free and start chatting with AI safely. Pro features are in active development — join the waitlist to be notified at launch.