Enterprise PII Guardrails Studio | Zero-Leak AI Privacy Gateway

Live Architecture Simulation

Zero-Leak Proxy Pipeline in Motion

Your AI Client App

Origin

LangChain / Python / Node

Sends prompt with SSN & name.

Status: Active

POST /mask

PII Gateway

<25ms

Sub-25ms Privacy Engine

SQLCipher AES-256

LRU Auth

Waiting

30+ Recognizers

Waiting

AES-256

Standby

Zero-Leak Payload

Cloud AI

Public Internet

OpenAI / Claude / Gemini

Sees irreversible tokens (<PERSON_1>).

Compliance: Zero Exposure

PHASE 1 OF 5 T = 0.0ms

Client Ingress: Dispatching Sensitive Request

The client application initiates an API call to the gateway containing a sensitive patient name and SSN.

ENDPOINT POST /mask

BOUNDARY On-Prem VPC

PII RISK Raw Unsanitized

HTTP Request Payload (POST /mask) application/json

{
  "text": "Patient Alice Walker (SSN: 987-65-4321) prescribed Amoxicillin."
}

Interactive Engine Sandbox

Experience Real-Time PII Scrubbing

Test live detection right now. Pick an enterprise preset or type sensitive text to see tokenization pills and bi-directional unmasking in action.

Raw Sensitive Prompt (Before Gateway) 138 chars

Simulate sub-25ms gateway interception

Sanitized Payload (Sent to OpenAI/Claude)

Latency 16.4 ms

Detected 3 Entities

Vault Security AES-256

Modern Management Portal

The Studio in Action

Inspect telemetry, manage hardware-locked licenses, audit decrypted reversible vaults, and configure entity policies with sub-25ms response time.

http://localhost:8000/ui/index.html

Live Playground View

Visual badge pills pinpoint exactly what the engine detected before it ever leaves your network.

Developer Quickstart

Up & Running in 2 Simple Steps

Deploy the local air-gapped gateway, then connect your application in 3 lines of code.

Spin Up Privacy Gateway Server Air-Gapped / Offline

Launches high-performance reverse privacy proxy on http://localhost:8000.

bash - pip distribution

BASH

Studio dashboard initializes at → http://localhost:8000/ui/

Integrate Privacy Endpoints (Mask & Unmask) <25ms Bi-Directional

Scrub PII before sending to Cloud LLMs (POST /mask), then restore tokens (POST /unmask).

python - openai / cloud llm integration

PYTHON

Zero-Leak Model Ecosystem:

OpenAI Anthropic Claude Google Gemini LangChain Ollama

Cryptographic Privacy Pipeline

Zero-Trust Gateway Architecture

Raw sensitive customer records never leave your network boundary. Here is how our bi-directional privacy proxy secures your LLM prompts:

Client / App Input

App sends prompt containing names, SSNs, credit cards, or internal IP addresses to POST /mask.

"Alice SSN 123-45..."

LRU Auth Cache

X-API-Key verified in-memory in under 0.1ms with zero database lock contention.

Auth verified <0.1ms

Multi-Tier Recognizer

30+ entity recognizers + custom regex scrub data. Reversible tokens saved asynchronously to SQLCipher AES-256.

SQLCipher AES-256 Vault

External LLM Query

OpenAI, Claude, or Gemini receives only sanitized pseudonymized tokens (<PERSON_1>).

Zero Sensitive PII Leak

Authorized Unmask

POST /unmask seamlessly restores tokens in model completion back to real values for authorized staff.

Original Values Restored

Defense-in-Depth

Engineered for Strict Enterprise Compliance

Built from the ground up for healthcare, fintech, insurance, and defense organizations where cloud phone-home is strictly prohibited.

Sub-25ms Live Masking

In-memory LRU authorization cache and asynchronous non-blocking SQLite worker thread guarantee minimal latency impact on production streaming pipelines.

SQLCipher AES-256 Storage

Zero data at rest is left unencrypted. Database file, reversible token maps, API keys, and audit logs are encrypted using 256-bit AES-CBC encryption.

30+ Entity Recognizers

Pre-configured for SSN, credit cards, IBAN, passports, medical IDs, phone numbers, IP/MAC addresses, API keys, JWTs, and custom enterprise regex patterns.

Ed25519 Node-Locked Licensing

Commercial licenses are cryptographically locked to host Server Hardware IDs via Ed25519 asymmetric signatures, operating 100% offline with zero cloud phone-home.

Real-Time Threat Intelligence

Integrated firewall blocking malicious IP addresses, phishing domains, and credential leaks before prompt evaluation, protecting internal LLMs from poison vectors.

Bi-Directional Unmasking

Effortlessly restore LLM completions back to original values for authorized human agents, ensuring end users receive normal responses while AI models see zero PII.

Built For Global Regulatory Compliance

HIPAA Security Rule Section 164.312 GDPR Article 32 Pseudonymization PCI-DSS v4.0 Primary Account Protection SOC 2 Type II Readiness

🎁 Community Launch Special

6 Months Free Enterprise Evaluation

To celebrate our v2.0 release, activate a full Enterprise Tier key with unlimited requests, unrestricted payload length, and all 30+ entity recognizers, valid through March 31, 2027.

1. Launch Studio Open dashboard at http://localhost:8000.

2. Open Settings Navigate to Settings → License & Quota tab.

3. Paste & Activate Paste product key and enjoy 6 months unlimited.

Enterprise Tier Feature Matrix

Transparent capabilities across trial, standard, and enterprise deployments.

Capability	Trial Tier	Community Promo	Standard Tier	Enterprise Tier
Masking Latency	< 25ms	< 25ms	< 25ms	< 25ms
Entity Recognizers	30+ Included	30+ Included	30+ Included	30+ Included
Custom Regex Engine	1 Pattern	Unlimited	5 Patterns	Unlimited
Max Payload Length	5,000 chars	Unrestricted	50,000 chars	Unrestricted
Monthly Quota	1,000 reqs	Unlimited	20,000 reqs	Unlimited
Storage Security	SQLCipher AES-256	SQLCipher AES-256	SQLCipher AES-256	SQLCipher AES-256
Offline Node-Locking	Standard Floating	Standard Floating	Standard Floating	Ed25519 HWID Lock
License Term	Permanent Free	6 Months Free	Commercial Key	Perpetual Node-Locked