◆ LocalMask

AI reads your code,
not your secrets.

Mask an entire repo or a single file — LocalMask replaces every secret, API key, and PII with a reversible token before anything reaches Claude, GPT, or Copilot. Two workflows, one tool, 100% local.

Install free CLI VS Code Cursor / Windsurf JetBrains
See both workflows ↓ Watch demo ↓ Pro plans →
pip install localmask click to copy

Free, open source, runs offline. Try the interactive demo below →

Local-first privacy for AI coding

This is what the AI
actually sees.

Scan the repo, mask credentials and PII into reversible tokens, and send only tokens to Claude, GPT, or Gemini. Answers come back and rehydrate on your machine. Try the sensitivity levels on the real .env file →

No spam — just product updates and new features.

You're on the list!
Sensitivity level
payments-service / .env scanning… 8 secrets masked · safe to send 
# payments-service/.env — sent to ClaudeDATABASE_URL=postgres://payments_rw~[DB_USER_0]~:pV9$kQ2!zR~[PASSWORD_0]~@db-prod-01.acme.internal~[HOST_0]~:5432~[PORT_0]~/payments~[DB_NAME_0]~STRIPE_SECRET_KEY=sk_live_51MspeZv8Klo2CqR7xY~[API_KEY_0]~JWT_SIGNING_KEY=hs256-9f3a7c1e8b2d4061~[API_KEY_1]~AWS_ACCESS_KEY_ID=AKIA5J7QX9P2M4RTUVWX~[AWS_KEY_0]~ONCALL_EMAIL=[email protected]~[EMAIL_0]~INTERNAL_API=https://vault.acme.internal~[HOST_1]~/v2
real secret becomes → token the AI sees
See it in action

Watch how LocalMask works.

Pick a demo. Each one is under 3 minutes.

Full Repo Workflow 2:45

Scan a repo, review detections, publish a masked mirror, ask AI safely.

0:00 / 0:00
VS Code Prompt Builder 1:50

Mask any file in VS Code, teach missed secrets, copy a safe prompt to any AI, rehydrate the answer.

0:00 / 0:00
Two ways to mask

Full repo or single file — you choose the workflow.

LocalMask protects your code in two ways. Use both, or pick the one that fits your moment.

Repo mirror

Scan a repo, publish a masked copy

LocalMask scans your entire repository and creates a private masked mirror. The AI reads the mirror — your real secrets stay on your machine.

  1. Scan — point LocalMask at any Git repo
    localmask scan .
  2. Publish — create a masked mirror repo
    localmask publish --mirror
  3. Sync — keep the mirror up to date as you code
    localmask sync
  4. AI reads the mirror — every secret is a stable ~[TOKEN]~

Best for: letting AI agents, Copilot, or Claude Code read your full codebase continuously — without ever exposing secrets.

Prompt Builder

Mask one file, copy-paste to any AI

Open any file in VS Code — no repo scan needed. The Prompt Builder masks it instantly, gives you a ready prompt, and rehydrates the AI's answer.

  1. Open — hit Cmd+Alt+M on any file
  2. Masked editor — see the masked version side-by-side with a prompt panel
  3. Copy prompt — pick a template, add your question, copy the masked prompt
  4. Paste in AI — ChatGPT, Claude, Gemini — any chat window
  5. Rehydrate — paste the AI's answer back, tokens become real values

Best for: quick questions about a config, migration, or sensitive file — no setup, no repo, just mask and ask.

What you get

Everything you need to keep secrets out of AI.

LocalMask doesn't decide what may reach the AI — it helps make what reaches the AI safe to send. And with your reviews and teaching, it keeps getting better.

Masked repo mirror

  • Scan a repo, publish a private masked copy the AI reads
  • Every secret becomes a stable ~[TOKEN]~
  • Git sync keeps the mirror current
  • Works with GitHub, GitLab, Bitbucket, self-hosted

Prompt Builder

  • Cmd+Alt+M on any file — no repo needed
  • Masked editor + prompt panel side by side
  • Pick a template, add your question, copy
  • Paste the AI answer back → rehydrate tokens to real values
  • Teach the model: right-click → mark as secret

Local, learning, yours

  • Persistent encrypted vault — stable tokens across restarts
  • Editable detection rules, no code
  • Finds secrets, PII, and customer data: DOB, CVV, passport, national ID, GPS, VIN, license plates, and more
  • 9 language packs — national IDs, customer & booking IDs, checksum-validated

Pro & Team extras

  • AI proxy — masks prompts before they reach your AI / gateway (Pro)
  • Local AI model that learns from your reviews (Pro)
  • Web dashboard (Pro)
  • Team-shared vault, LDAP/AD, SSO, audit trail (Team)
See pricing →

Free & open source on GitHub ↗ — run it entirely offline with the CLI + MCP server.

Free IDE extensions

Mask, build prompts, and review — without leaving your editor.

Toggle real ↔ masked view, open the Prompt Builder on any file (Cmd+Alt+M), review detections inline, and teach the model — all from your editor. No repo scan required for the Prompt Builder. The extensions are free and open source.

VS Code
Marketplace
Cursor · Windsurf
OpenVSX
JetBrains
IntelliJ · PyCharm · WebStorm

Also works as a CLI + MCP server — no IDE required. See pricing →

Pricing

Free forever. Pro is one-time. Team & Enterprise are annual, per seat.

Free · OSS

$0
open source
  • Regex + entropy engine
  • Finance Mode: mask money amounts
  • 9 language packs
  • Masked-repo mirror the AI reads
  • CLI + MCP server
  • Edit detections
Get the CLI

Pro

€49
one-time · 12 months of updates · +VAT
  • Everything in Free
  • Local AI model + learning
  • Web dashboard
  • AI proxy — masking before your AI / gateway
  • Finance Mode: you choose what the AI sees
Buy Pro

Team

€199
per seat / year · +VAT
  • Everything in Pro
  • Shared org rules + seat management
  • LDAP/AD + SSO sign-in
  • Tamper-evident audit trail
  • Closed environment — lock all egress
  • Lock security settings org-wide
  • Priority support
Buy Team

Enterprise

Custom
let's talk
  • Everything in Team — same features
  • Unlimited seats
  • Dedicated support & SLA
  • Procurement, invoicing, DPA
Contact sales

✓ 14-day money-back guarantee — no questions asked.

Pro is a one-time purchase: every version released within your 12-month update window is yours to keep and run forever. Validated offline — no phone-home, no monthly re-activation. Team & Enterprise are annual, per seat.

By purchasing you agree to our License Terms · Refund Policy · Privacy.

Palantir CEO on CNBC · July 2026

Are you keeping the data?
LocalMask makes the question irrelevant.

Alex Karp told CNBC that enterprises fear one thing about AI vendors: they pay for tokens and hand over their IP. He's right about the worry. Here's the architecture that removes it.

The worry — watch the clip

Karp says something has gone completely wrong with how AI is sold: enterprises send their data and their alpha to model vendors and can't answer the most basic governance question — who ends up holding it?

Clip: CNBC Squawk Box, July 1 2026 (official CNBC YouTube channel)

The answer — mask before it ever leaves

sk-live-4f9d… → ‹API_KEY_01› // the real value never leaves your machine
  • Secrets and PII are detected and masked locally, before any prompt reaches Claude, GPT, or any model.
  • Only metadata travels — file path, line number, pattern type. Never the value.
  • The token map stays on your disk, encrypted, git-ignored. No phone-home. Ever.
  • Keep using frontier AI. Stop shipping your IP with every prompt.
See how the pre-gate works

LocalMask Pro is not affiliated with, or endorsed by, Palantir, Alex Karp, or CNBC. The clip is embedded from CNBC's official YouTube channel and quoted for commentary.

Get started

Lock it down before you ship it to AI.

Free, open source, runs entirely on your machine. Install in one command and scan your first repo in under a minute.

pip install localmask
Download tarball Read the docs

Want more? Pro is a one-time €49 — adds the AI proxy, learning model, and dashboard.

Get notified about updates and new features:

You're on the list. We'll reach out when the beta opens.