1Enter / paste text
Enter text that contains sensitive information and press “Mask.” Detected items are replaced with tokens like [NAME_1].
Uses a Japanese morphological dictionary to also catch names without honorifics and proper nouns not in the built-in list. Turning it on downloads the analysis dictionary (~17 MB) from this site (same origin) on first use only. Your input text is never sent.
2Masked result
Mapping table (token ↔ original)
Stored in your browser (IndexedDB) · 0 items| Token | Original (sensitive) | Category |
|---|
3Restore (demask) — turn the AI's reply back
Paste an AI reply that still contains the tokens, and it is instantly restored to the original terms using the mapping table.
Process whole folders / files locally
Load all text files in a folder at once and mask them in parallel inside your browser. The originals are left untouched and saved as name_masked.ext. Everything stays on your device — nothing is uploaded.
“Open folder” writes directly into the same folder on Chrome/Edge (File System Access API). On unsupported browsers or with “Select files,” it falls back to download saving.
Toggle detection categories
Switch detection targets to fit your use case. Settings are saved on this device (localStorage).
Custom term list
Register your own must-mask terms such as project names or internal codes.
Type term=alias to set a custom token (e.g. Orion=PROJECT_X).
Privacy settings
Reliably erases the in-browser mapping data when the session ends.
01What this tool does
Pasting company names, personal names, contacts, or API keys straight into external AIs like ChatGPT or Claude is risky. Local Mask automatically converts them into tokens before sending, then restores them in one click after you receive the AI's reply. Detection, conversion, and restoration all happen only inside your browser — your input is never sent anywhere.
02Basic usage (round-trip clipboard)
Paste text containing secrets and press “Mask.” Names become [NAME_1], emails become [EMAIL_1], and so on.
Use “Copy masked text,” paste it into the AI chat, and ask your question. The AI never receives the real information.
Paste the AI reply that still contains the tokens into the “Restore” box and run it; the original terms come back instantly via the mapping table.
When finished, use “Discard entire mapping.” Auto-discard on closing the tab is also on by default.
03Process files in bulk
On the “Batch files” tab, open a folder of logs etc. or drag files in.
All files are processed together. On Chrome/Edge they are saved directly into the same folder with a _masked suffix; otherwise via download.
04Main detected information
- Contacts & location: email addresses, phone numbers, addresses, postal codes (regex + dictionary)
- Secrets & IT: credit card numbers (Luhn check), API keys/tokens (entropy analysis), passwords (label detection)
- People & orgs: personal names, company/organization names (heuristics using honorifics, titles, legal forms, surname dictionary)
- Custom: your own terms registered under “Detection settings”
05Frequently asked questions
QIs my entered information sent anywhere?
QWhere is the mapping table stored?
IndexedDB and used for demasking (restoration). You can erase it with “Discard entire mapping” or the default-on “Auto-discard when the tab closes.”