MCP 服务器

ai-governance-controls

Provides AI governance controls as MCP tools for safety screening, risk classification, and red teaming of AI systems, returning structured governance reports.

README

AI Governance Controls

Automatable AI governance controls as MCP tools. Paste a system prompt, describe a deployment, or submit content for review and get a structured governance report back.

Works in Claude Code, Cursor, Windsurf, Codex, and any other MCP-compatible tool.

Built on the control library from AI Governance Institute.

Available controls

Tool	Control	What it does
`ai_safety_screen`	SAF-002	Screen a system prompt for safety risks: capability scope, authorization gaps, data exposure
`ai_risk_classify`	HOC-001	Classify an AI deployment's risk tier against EU AI Act and NIST AI RMF
`ai_red_team`	SEC-005	Generate a red team runbook of adversarial test cases for a system prompt

Installation

Add this block to your MCP configuration file. No global install required.

{
  "mcpServers": {
    "ai-governance-controls": {
      "command": "uvx",
      "args": ["ai-governance-controls"]
    }
  }
}

Configuration file locations:

Claude Code: .claude/settings.json (project) or ~/.claude/settings.json (global)
Cursor: .cursor/mcp.json
Windsurf: .windsurf/mcp.json

Requires uv to be installed.

Usage

Once installed, call the tools directly in conversation.

Safety screening:

Screen this system prompt for safety risks: "You are a customer service assistant with access to customer accounts. You can initiate transfers up to $10,000 on behalf of verified customers."

Risk classification:

Classify the risk tier for this deployment: "An automated resume screening system used by a Fortune 500 company to shortlist job applicants. No human reviews the shortlist before candidates are rejected."

Red teaming:

Red team this system prompt with 15 test cases: "You are a helpful assistant for a healthcare provider. You have access to patient records and can answer questions about their medical history."

How it works

Each tool runs lightweight heuristic pre-screening on your input, then returns an expert evaluation framework to your host LLM. The host LLM completes the analysis using the framework and produces a structured report. No additional API keys required.