KitWharfFREEFast & Private Web Apps
Zero-Knowledge · Reversible

AI Prompt Sanitizer & Scrubber

Scrub API keys, internal IPs, hostnames, local directory paths & PII locally in browser RAM before pasting into ChatGPT, Claude, or Gemini — then auto-restore tokens in the AI reply.

100% Local RAM · Zero UploadsScrubbed 7 matches (7 unique) in 3.5ms · 0 bytes sent
Token Mode:
Presets:
acme-corp

1. Raw Input (Never Leaves Browser)

567 chars · 5 lines
Safe to paste into ChatGPT, Claude, Gemini, or Cursor.

Session Token Mapping Vault (7 unique items)

Toggle any item off if you want to keep it unmasked (e.g. public error codes or non-sensitive strings).

CategoryDetected TypeReplacement TokenOriginal ValueHitsAction
NetworkIPv4 Address{{IP_1}}10.••••••941
NetworkInternal Hostname{{HOST_1}}api••••••al1
NetworkIPv4 Address{{IP_2}}10.••••••121
SecretsJWT Token{{JWT_1}}eyJ••••••Xk1
UUIDsUUID{{UUID_1}}550••••••001
NetworkInternal Hostname{{HOST_2}}pos••••••al1
NetworkIPv4 Address{{IP_3}}192••••••451
Coming Soon: KitWharf Pro

Want Zero-Switch AI Troubleshooting, Multi-File .har/.env Bundles & Chrome Extension Sync?

The core web scrubber above is free forever. KitWharf Pro will add a BYOK & Local Ollama split-pane AI debugger with live auto-restore, AES-256-GCM encrypted company rulebooks, and an inline Chrome/Edge extension for ChatGPT, Claude & Gemini.

Workflow Guide

How to Sanitize Prompts & Restore AI Replies Safely

1

Paste Logs or Scripts

Paste stack traces, .env files, PowerShell/Bash scripts, or drop a .har/.log file directly into browser RAM.

2

Inspect & Tune Masks

Toggle categories (Secrets, Network, Paths, PII), switch between Bracket or Format-Preserving tokens, and add custom company terms.

3

Ask Any AI Safely

Copy the sanitized prompt into ChatGPT, Claude, Gemini, or Cursor without exposing credentials, internal IPs, or local usernames.

4

Auto-Restore Reply

Paste the AI's fix into Restore AI Reply to swap all tokens back to your real production IPs, paths, and variables locally.

Zero-Knowledge Browser RAM

No backend API, no prompt logging, and no telemetry on your text. Even on corporate-locked laptops where extensions are blocked, scrubbing stays 100% local.

Format-Preserving Synthesis

Private IPs map deterministically into RFC 5737 test ranges (192.0.2.x) while preserving /24 grouping, and scripts keep valid Windows/Unix directory structure.

Collision-Safe Restoration

Identical secrets or paths share one deterministic token across your entire log, and longest-match-first restoration unmasks AI fixes cleanly without partial overlaps.

Detection Engine Reference

What Gets Scrubbed & How Tokens Preserve Context

Every category below executes 100% locally in a single pass with deterministic deduplication and reversible restoration.

CategoryExamples DetectedBracket TokenFormat-Preserving Token
API Keys & SecretsAWS AKIA..., OpenAI/Anthropic sk-..., GitHub PATs, Stripe, JWTs, PEM keys, DB URIs{{SECRET_1}}[REDACTED_SECRET_1]
Network & InfraIPv4/CIDR (10.42.18.94), MACs, internal hosts (*.internal, *.local, *.corp){{IP_1}} / {{HOST_1}}192.0.2.1 / host-1.example.internal
Directories & PathsWindows (C:\Users\name\repo\...), UNC shares, Unix (/home/..., /var/www/...){{PATH_1}}C:\workspace\dir-1 / /workspace/dir-1
PII & NamesEmails, phones, Luhn-checked cards, SSNs, labeled & email-linked person names{{EMAIL_1}} / {{NAME_1}}user-1@example.invalid / Alex Morgan
Config & EntropyPASSWORD=, TOKEN=, WEBHOOK_SECRET:, high-entropy strings{{CRED_1}}[REDACTED_CRED_1]

DevOps & SRE Incident Triage

Sanitize Nginx, Kubernetes, and PostgreSQL logs with Format-Preserving RFC 5737 IPs so AI still traces upstream timeouts without learning your private VPC topology.

PowerShell & Bash Script Debugging

Automatically mask C:\Users\... and /home/... project paths alongside embedded tokens before asking an LLM to refactor deployment scripts.

Support Tickets & Compliance

Draft polite customer replies with ChatGPT or Claude while keeping customer names, emails, phone numbers, and card digits strictly inside your browser.

Security Benchmark

KitWharf AI Prompt Sanitizer vs. Manual Redaction vs. Cloud Proxies

CapabilityKitWharf Prompt SanitizerManual Find & ReplaceCloud DLP Proxies
Zero-Upload Execution100% In-Browser RAMLocal (Error-Prone)Routed Through 3rd-Party Servers
Reversible AI Reply RestoreYes (1-Click Auto-Restore)ManualEnterprise Only
Directory Paths & RFC 5737 IPsYes (Format-Preserving)NoPartial
No Install / No Extension RequiredInstant Web AppYesRequires MDM / Proxy Setup

Frequently Asked Questions

Everything you need to know about sanitizing logs, scripts, API keys, and PII locally before using AI.

No. All pattern matching, Shannon entropy scoring, Luhn credit card validation, token replacement, and AI reply restoration execute 100% locally inside your browser's JavaScript memory with zero server uploads.

Explore More Utilities

Complementary Tools

Other fast, free, and zero-upload utilities built for your workflow.

View all 87 utilities