Skip to content

Releases: moxno/privacyscrubber-mcp

v2.2.3: Zero-Trust MCP Server & Lightweight Node.js Presidio Alternative SDK

Choose a tag to compare

@moxno moxno released this 20 Sep 21:46

🚀 PrivacyScrubber v2.2.3: Zero-Trust MCP Server & The Node.js Presidio Alternative

This release brings full parity across all 15 Model Context Protocol (MCP) tools and introduces the official Lightweight, Zero-Dependency Presidio Alternative for Node.js / TypeScript backends and RAG pipelines.

🌟 What's New in v2.2.3

  1. 🥊 The Node.js Presidio Alternative ():

    • Zero dependencies (~150KB runtime footprint vs Presidio's 500MB+ Python/spaCy daemon).
    • Sub-millisecond in-memory latency (<1ms) with zero network egress.
    • 1-line transparent drop-in wrapper for OpenAI (), Express middleware, and LangChain transforms.
    • Real-time streaming de-tokenization via native .
  2. 🛡️ 15 Production-Grade MCP Tools:

    • : Zero-Trust shell execution wrapper intercepting secrets before LLM receipt.
    • : Safe config/secrets file reader.
    • & : Tokenized agentic diff review with authentic local secret restoration on disk.
    • : Instant generation of .cursorrules, CLAUDE.md, and .windsurfrules.
  3. 🔒 Hardened Intellectual Property & Parity:

    • AST-obfuscated core engine deployed to npm with Base64 string arrays.
    • 25 compliance profiles (HIPAA, GDPR, SOC 2, PCI DSS, Legal, Dev, Finance).
    • Zero data loss: reversible deterministic tokens ([NAME_1], [EMAIL_1], [AWS_KEY_1]).

📦 Quickstart

MCP Server (Cursor, Claude Desktop, Windsurf):

npx -y @privacyscrubber/mcp-server

Programmatic SDK (Node.js, Next.js, RAG Pipelines):

npm install @privacyscrubber/sdk
import OpenAI from 'openai';
import { wrapOpenAI } from '@privacyscrubber/sdk';

const openai = wrapOpenAI(new OpenAI({ apiKey: process.env.OPENAI_API_KEY }));

v2.2.0: Defense-in-Depth PII Redaction, Trust Boundary Specification & Multi-OS CI

Choose a tag to compare

@moxno moxno released this 14 Sep 18:43

🛡️ PrivacyScrubber MCP Server v2.2.0

Highlights

  • Trust Boundary & Threat Model Specification (SECURITY_MODEL.md):
    • Formal documentation of in-process defense-in-depth sanitization vs network airlocks.
    • Transparent analysis of the reverse-unmasking (de-tokenization) lifecycle and operational rules for terminal boundary output.
    • Host execution boundary specification for guard_exec and filesystem tools with recommended container isolation guidelines.
  • Automated Multi-OS CI Pipeline:
    • Continuous integration across Ubuntu and macOS on Node.js 18.x, 20.x, and 22.x.
    • Full suite verification of DLP regex profiles, offline document parsing, and agentic guard tools.
  • Agentic Guard Protocol & CLI:
    • Integrated ps-guard / pii-masking-run CLI for UNIX pipelines (cat file | ps-guard).
    • Automated agent instruction generation for Cursor (.cursorrules), Windsurf (.windsurfrules), Claude (CLAUDE.md), Copilot, and Cline.
  • 100% On-Device Zero Egress:
    • Volatile heap RAM-only session maps, zero persistence, zero telemetry, and verifiable offline operation (Airplane Mode).
    • Open-source under the MIT License.