Caveman.PrivacyGuard.Mcp
1.3.0
dotnet add package Caveman.PrivacyGuard.Mcp --version 1.3.0
NuGet\Install-Package Caveman.PrivacyGuard.Mcp -Version 1.3.0
<PackageReference Include="Caveman.PrivacyGuard.Mcp" Version="1.3.0" />
<PackageVersion Include="Caveman.PrivacyGuard.Mcp" Version="1.3.0" />
<PackageReference Include="Caveman.PrivacyGuard.Mcp" />
paket add Caveman.PrivacyGuard.Mcp --version 1.3.0
#r "nuget: Caveman.PrivacyGuard.Mcp, 1.3.0"
#:package Caveman.PrivacyGuard.Mcp@1.3.0
#addin nuget:?package=Caveman.PrivacyGuard.Mcp&version=1.3.0
#tool nuget:?package=Caveman.PrivacyGuard.Mcp&version=1.3.0
Caveman.PrivacyGuard.Mcp
MCP (Model Context Protocol) server for Caveman.PrivacyGuard — exposes PII detection/masking, session-based restore, AI-transparency notices, and prompt-injection screening as MCP tools for Claude Code, OpenCode, Cursor, and any MCP-compatible agent.
Tools
| Tool | Description |
|---|---|
analyze_text |
Detects and scores PII/sensitive data (32 countries), returns compliance flags (GDPR/AI Act/NIS2/PCI-DSS/NIST) |
mask_text |
Masks sensitive data; pass a sessionId to get unique [PG_N] placeholders restorable later |
restore_text |
Restores original values in a model response, given the sessionId used to mask it |
check_prompt_injection |
Heuristically screens untrusted text for prompt-injection attempts before it reaches an LLM |
get_ai_transparency_notice |
Returns a configurable, localized disclosure that the user is talking to an AI system |
Usage
Add to your MCP client config (e.g. Claude Code, Cursor):
{
"mcpServers": {
"caveman-privacyguard": {
"command": "dotnet",
"args": ["path/to/caveman-privacyguard-mcp.dll"]
}
}
}
Each conversation should use a distinct sessionId when calling mask_text / restore_text, so different conversations never share (or leak into) each other's placeholder mappings — see PrivacySessionManager in the core library.
Disclaimer
analyze_text, check_prompt_injection, and get_ai_transparency_notice are technical aids, not a compliance or security guarantee. GDPR, AI Act, and NIS2 obligations require contextual legal assessment; prompt-injection screening is heuristic and should be one layer among several defenses (system-prompt hardening, output filtering, least-privilege tool access).
License
MIT — see LICENSE.
| Product | Versions Compatible and additional computed target framework versions. |
|---|---|
| .NET | net8.0 is compatible. net8.0-android was computed. net8.0-browser was computed. net8.0-ios was computed. net8.0-maccatalyst was computed. net8.0-macos was computed. net8.0-tvos was computed. net8.0-windows was computed. net9.0 was computed. net9.0-android was computed. net9.0-browser was computed. net9.0-ios was computed. net9.0-maccatalyst was computed. net9.0-macos was computed. net9.0-tvos was computed. net9.0-windows was computed. net10.0 was computed. net10.0-android was computed. net10.0-browser was computed. net10.0-ios was computed. net10.0-maccatalyst was computed. net10.0-macos was computed. net10.0-tvos was computed. net10.0-windows was computed. |
-
net8.0
- Caveman.PrivacyGuard (>= 1.3.0)
- Microsoft.Extensions.Hosting (>= 8.0.1)
- ModelContextProtocol (>= 1.4.0)
NuGet packages
This package is not used by any NuGet packages.
GitHub repositories
This package is not used by any popular GitHub repositories.
1.3.0: aligned with Caveman.PrivacyGuard 1.3.0. New tools: mask_stream (boundary-safe streaming PII masking over an array of text chunks, with a configurable overlap window), mask_document (streaming masking of plain-text/Markdown/CSV files into a name.masked.ext copy, original never overwritten), and session_summary (redacted placeholder+category view of a session — never the original value). 1.2.2 added 5 new country validators, AI Act/NIS2 compliance flags, PrivacySessionManager, AiTransparencyNotice and ChatIntentRouter.