🔒 100% local · Open source · Free

Paste into AI.
Keep the secrets.

PasteGuard masks emails, phone numbers, card numbers, API keys and client names the moment you paste into ChatGPT, Claude or Gemini. Then it shows the real values back to you in the AI's reply.

✓ No account✓ No servers. Nothing is uploaded.✓ Works on ChatGPT, Claude, Gemini, Copilot, Perplexity
Live demo

See exactly what the AI gets

Edit the text on the left. This page runs the same detection engine as the extension, right in your browser.

📋 What you paste
➜
🤖 What the AI sees

The AI replies using placeholders…

💬 AI reply, real values restored in your browser
Click “Simulate AI reply”.
How it works

Three things happen in under a millisecond

01 · Catch

You paste as usual

PasteGuard checks the clipboard text before the chat box receives it. No new habit, no extra click.

02 · Mask

Sensitive values become placeholders

[EMAIL_1], [CLIENT_1]… The same value always gets the same placeholder, so the AI's answer still makes sense.

03 · Restore

You read the real values

When the reply mentions a placeholder, PasteGuard swaps the real value back on your screen only. Copy the reply and you get the real values too.

What it catches

Understands context, not just patterns

Card numbers are Luhn-checked and IBANs and Aadhaar numbers checksum-validated. Context rules catch what patterns can't: “CVV 123”, “client ID HLX-00482”, “finance officer Rohan Deshmukh”. Once a name or company is found, every later mention is masked too. Order numbers, dates and version numbers are left alone.

Privacy by design

A privacy tool that doesn't need your trust

  • ●
    No servers. Detection runs in your browser. There's no backend to send anything to.
  • ●
    Memory-only vault. The placeholder-to-value map is wiped when you close the browser.
  • ●
    Minimal permissions. Only storage, and it only runs on AI chat sites.
  • ●
    Open source. Read every line on GitHub.
// manifest.json: the entire permission list "permissions": ["storage"], "content_scripts": [{ "matches": [ "https://chatgpt.com/*", "https://claude.ai/*", "https://gemini.google.com/*", … ]}] // no host permissions, no network calls, // no analytics, no remote code
Install

Up and running in a minute

Chrome Web Store ● Live

One click to install. Updates arrive automatically. Works in Chrome, Edge, Brave and Arc.

Add to Chrome, it's free

After installing, pin the 🛡️ icon and paste something into ChatGPT to see it work.

Manual install (for developers)

  1. Download the .zip and unzip it.
  2. Open chrome://extensions and switch on Developer mode (top right).
  3. Click Load unpacked and choose the unzipped pasteguard folder.
  4. Pin the 🛡️ icon. Paste something into ChatGPT to see it work.
Pricing

Free for individuals. Always.

Free
$0
Everything you need on your own
  • 25 built-in detectors with context awareness
  • Custom "always mask" terms
  • Real values restored in replies
  • 9 AI chat sites
Add to Chrome
Pro Coming soon
$5/month
For consultants, lawyers, recruiters
  • Mask typed text, not just pastes
  • File uploads (PDF, DOCX) scanned before upload
  • Saved term lists per client
  • Any website, not just AI chats
⭐ Star on GitHub to follow
Team Coming soon
$6/user/month
For companies adopting AI safely
  • Company-wide term lists and rules
  • Dashboard: what was blocked, where
  • Audit log for compliance
  • Deploy via Google Workspace / Intune
Request early access
FAQ

Questions

Does PasteGuard send my data anywhere?

No. There's no server. The extension has no network permissions beyond the AI sites it runs on, and it makes no requests of its own.

What if I actually want to paste something unmasked?

Click “Allow next paste unmasked” in the notification and paste again, or switch PasteGuard off from the toolbar icon.

Does it catch names?

Yes, from context: titles (“Dr. Ananya Rao”), roles (“finance officer Rohan Deshmukh”), labels (“Name: …”), greetings, common first names, and names implied by email addresses. Company names are caught from suffixes (“Inc”, “Labs”, “Biosciences”) and email domains. Once found, every later mention is masked. For anything unusual, add it to “Always mask”.

Does it mask text I type?

Version 0.1 masks pastes, which is where bulk sensitive data comes from (emails, tickets, spreadsheets). Typed-text scanning is planned.

Is it a replacement for my company's security tools?

No. It's a seatbelt for everyday use. It catches common sensitive values but can't guarantee it catches everything.