Governmyai

@governmyai/mcp-server

GovernMy.ai MCP server — query AI governance obligations from Claude Desktop, Cursor, or any MCP-compatible client.

Covers EU AI Act, ISO 42001, Colorado AI Act, NIST AI RMF, HIPAA (AI provisions), SOX (AI provisions), and FTC AI guidance.

What this does

This MCP server gives an AI assistant (like Claude) the ability to query GovernMy's rules engine directly while you're working. You can ask questions like:

• "What EU AI Act obligations apply if I build a hiring AI for the EU?"
• "Which frameworks does this cover?"
• "What evidence do I need for EU AI Act Article 10 data governance?"
• "Does my ISO 42001 Clause 6 work also cover any NIST AI RMF functions?"

...and get structured regulatory obligations back as part of the conversation.

Important: this tool returns obligations, not verdicts. It never tells you "you're compliant" or "you're not compliant" — those are determinations only a human reviewer can make. Every response includes a humanReviewRequired flag where applicable.

Copy-paste into Claude

After installing + restarting, try one of these:

• "What are the mandatory-review obligations for a high-risk hiring AI deployed in the EU?"
• "Show me the evidence requirements for EU AI Act Articles 11 and 14."
• "Compare EU AI Act and Colorado AI Act obligations for a credit-scoring AI."
• "Does my HR chatbot trigger any regulatory obligations? It does resume screening."
• "What ISO 42001 Annex A controls do I need for a generative AI system?"
• "Which NIST AI RMF dimensions does EU AI Act Art. 10 data governance also satisfy?"
• "I'm building a medical imaging AI — what obligations apply under HIPAA AI provisions?"
• "List every obligation flagged as cannotBeAutoSatisfied across all frameworks."

Install + configure (Claude Desktop)

1. Register for an API key at https://governmy.ai/developers (free tier available)
2. Open your Claude Desktop config file:
   • macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
   • Windows: %APPDATA%\Claude\claude_desktop_config.json
3. Add the GovernMy server to your mcpServers object:

{  "mcpServers": {    "governmy": {      "command": "npx",      "args": ["-y", "@governmyai/mcp-server"],      "env": {        "GOVERNMY_API_KEY": "ooa_live_your_key_here"      }    }  }}

4. Restart Claude Desktop. Ask Claude: "What governance frameworks does GovernMy cover?" — it should call the list_frameworks tool.

Install + configure (Cursor)

Add the server to your Cursor MCP settings (Settings → MCP Servers):

• Command: npx
• Args: -y @governmyai/mcp-server
• Environment: GOVERNMY_API_KEY=ooa_live_your_key_here

Configure (other MCP clients)

Any client that speaks stdio MCP can run this server. Set GOVERNMY_API_KEY in the client's env for the server process.

Environment variables

Tools exposed

• get_obligations — query obligations that apply to an AI system given context (risk tier, role, industry, etc.)
• classify_risk_tier — classify an AI system's EU AI Act risk tier (unacceptable / high / limited / minimal)
• list_frameworks — list the frameworks the engine covers with metadata
• get_rule — fetch the full payload for a single obligation by id
• get_evidence_requirements — fetch required evidence artifacts for a set of obligations
• get_cross_references — fetch cross-framework mappings for an obligation

Every tool response includes a notAVerdict notice reminding the AI assistant that obligations are not compliance verdicts.

What this is not

• Not legal advice. The rules engine is a research tool + workflow accelerator. Compliance decisions require qualified human review.
• Not a replacement for a compliance professional. If your AI system is materially exposed to any of the covered frameworks, pair this with a real compliance program.
• Not a verdict engine. We deliberately do not return "compliant: true" or "allowed: true". This is both an ethical line and a legal one (EU AI Act Art. 14).

Support

• Docs: https://docs.governmy.ai/mcp
• API reference: https://docs.governmy.ai/api
• Issues: https://github.com/governmyai/mcp-server/issues
• Email: [email protected]

License

MIT