@eucomplyhub/mcp-eu-ai-act 0.1.0

package/LICENSE

@@ -0,0 +1,27 @@
+MIT License
+
+Copyright (c) 2026 Piotr Reder / eucomplyhub.com
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+Risk classification logic in src/index.ts is adapted from
+@clustral/risk-compass (MIT) by Clustral AI Labs. Attribution preserved.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.
+
+This software is an educational tool for orientation, not legal advice.
+For binding compliance mapping, consult a qualified expert.

package/README.md

@@ -0,0 +1,213 @@
+# @eucomplyhub/mcp-eu-ai-act
+
+MCP (Model Context Protocol) server exposing free EU AI Act compliance classifiers to AI assistants. Built for **Claude Desktop**, **Cursor**, **Windsurf**, and any MCP-compatible client.
+
+> Powers AI assistants to evaluate mid-market SaaS systems against EU AI Act, NIST AI RMF, ISO/IEC 42001, OECD, GDPR, and sector-specific overlays.
+
+[![npm](https://img.shields.io/npm/v/@eucomplyhub/mcp-eu-ai-act.svg)](https://www.npmjs.com/package/@eucomplyhub/mcp-eu-ai-act)
+[![license](https://img.shields.io/npm/l/@eucomplyhub/mcp-eu-ai-act.svg)](LICENSE)
+
+---
+
+## Tools exposed
+
+### 1. `quick_risk_class` — 30-second multi-jurisdiction risk classifier
+
+Deterministic, rule-based AI risk classification. **No LLM call** — same inputs always produce the same outputs.
+
+**Frameworks covered:**
+- 🇪🇺 EU AI Act (Articles 5, 9–15, 26, 27, 50, 53)
+- 🇺🇸 NIST AI RMF (Govern · Map · Measure · Manage)
+- 🌐 ISO/IEC 42001 (Clauses 4–10 + Annex B)
+- 🌍 OECD AI Principles
+- 🇸🇬 Singapore Model AI Governance
+- 🇨🇳 PRC GenAI Interim Measures (for generation archetype)
+- 🔐 GDPR + UK GDPR + DPDP + CCPA + LGPD + PIPEDA (privacy stack)
+- 🏥 HIPAA / FDA SaMD / EU MDR-AI (healthcare overlay)
+- ⚖️ EEOC + NYC AEDT + Colorado SB 24-205 (US employment overlay)
+- 🏛️ OMB M-24-10 + CoE AI Convention (public sector overlay)
+
+**Inputs:**
+- `industry`: healthcare, publicSector, education, hr, retail, industrial, media, other
+- `archetype`: decisioning, generation, classification, recommendation, automation, forecasting
+- `impact`: internal, b2b, consumer, regulated
+
+**Returns:** Risk class (Critical / High / Limited / Minimal), 5-axis risk profile, per-framework verdicts.
+
+---
+
+### 2. `classify_annex3` — Deep Annex III classification (Claude-powered)
+
+Full EU AI Act mapping using Claude (Anthropic) with complete regulatory context — Articles 6, 9–15, 26, 27, 50, 53 plus the postponement nuance (Annex III standalone enforcement postponed to Dec 2 2027; Article 50 + GPAI Article 53 lock in Aug 2 2026).
+
+**Inputs:**
+- `company`: Company name
+- `industry`: Industry/vertical
+- `features`: Array of AI features
+- `useCase`: Plain-English description (min 20 chars)
+- `euExposure`: eu-customers-output, eu-employees-only, no-eu, considering-eu
+
+**Returns:**
+- Overall risk classification (high-risk / limited-risk / gpai / minimal-risk)
+- 8 Annex III categories with applies status + reasoning
+- Article 50 transparency obligations + reasoning
+- GPAI Article 53 applicability (provider / deployer) + reasoning
+- 5 priority remediation actions tailored to your stack
+
+> Calls https://eucomplyhub.com/api/annex3-classify — free, no signup, ~60s response time.
+
+---
+
+## Install
+
+### For Claude Desktop
+
+1. Install the package globally:
+
+   ```bash
+   npm install -g @eucomplyhub/mcp-eu-ai-act
+   ```
+
+2. Edit your Claude Desktop config file:
+
+   - macOS: `~/Library/Application Support/Claude/claude_desktop_config.json`
+   - Windows: `%APPDATA%\Claude\claude_desktop_config.json`
+
+3. Add the server:
+
+   ```json
+   {
+     "mcpServers": {
+       "eucomplyhub": {
+         "command": "npx",
+         "args": ["-y", "@eucomplyhub/mcp-eu-ai-act"]
+       }
+     }
+   }
+   ```
+
+4. Restart Claude Desktop.
+
+5. In a new chat, ask:
+
+   > Use the eucomplyhub tools to classify Acme Inc — an HR-tech SaaS with CV screening features for European enterprise customers.
+
+   Claude will call `classify_annex3` and return a structured Annex III mapping + priority remediation actions.
+
+### For Cursor
+
+Add to `.cursor/mcp.json` (workspace) or `~/.cursor/mcp.json` (global):
+
+```json
+{
+  "mcpServers": {
+    "eucomplyhub": {
+      "command": "npx",
+      "args": ["-y", "@eucomplyhub/mcp-eu-ai-act"]
+    }
+  }
+}
+```
+
+### For Windsurf
+
+Add to `~/.codeium/windsurf/mcp_config.json`:
+
+```json
+{
+  "mcpServers": {
+    "eucomplyhub": {
+      "command": "npx",
+      "args": ["-y", "@eucomplyhub/mcp-eu-ai-act"]
+    }
+  }
+}
+```
+
+### Generic stdio client
+
+```bash
+npx -y @eucomplyhub/mcp-eu-ai-act
+```
+
+The server runs over stdio (standard MCP transport).
+
+---
+
+## Example prompts
+
+Once installed, ask your AI assistant questions like:
+
+- *"What's the EU AI Act risk class for an HR-tech SaaS using AI to score job candidates?"*
+- *"Run the quick risk classifier for a healthcare diagnostic AI used by EU regulated medical providers."*
+- *"Deep-classify Acme Inc — they're a B2B fintech using AI to score loan applicants in the EU."*
+- *"For Notion AI features, what Annex III categories apply and what's the Article 50 obligation?"*
+
+The assistant will call the appropriate tool and return structured results you can act on.
+
+---
+
+## What does this cost?
+
+**Free.** Both tools call free public endpoints at `eucomplyhub.com`:
+
+- `quick_risk_class` runs entirely locally (deterministic JavaScript, no network call)
+- `classify_annex3` calls `https://eucomplyhub.com/api/annex3-classify` (rate-limited free public API)
+
+Rate limits apply for abuse prevention. Heavy usage should consider commissioning a full audit via [eucomplyhub.com/audit](https://eucomplyhub.com/audit).
+
+---
+
+## Disclaimer
+
+This MCP server is an **educational tool for orientation**, not legal advice. Each framework has specific clauses, exceptions, and edge cases. For binding compliance mapping (audit deliverables, certification prep, regulatory submission), consult an expert.
+
+For a paid expert audit:
+- 📋 **Tier 1 Quick Audit (€799):** Free /risk-class + /annex3 + 60-min consultation
+- 🔬 **Tier 2 Full Audit (€1,999):** Triple-framework methodology + audit-ready deliverable
+- 🛡️ **Tier 3 Continuous Monitoring (€299/mo):** Post-audit ongoing review
+
+Book at [eucomplyhub.com/audit](https://eucomplyhub.com/audit).
+
+---
+
+## Methodology
+
+Built and maintained by [Piotr Reder](https://www.linkedin.com/in/piotrreder/) ([eucomplyhub.com](https://eucomplyhub.com)). Triple-framework specialist for mid-market SaaS preparing for EU AI Act enforcement.
+
+**Risk class logic** adapted from [@clustral/risk-compass](https://github.com/king-star-12/risk-compass) (MIT).
+
+**Annex III deep classifier** powered by [Anthropic Claude](https://www.anthropic.com/) (claude-sonnet-4-6).
+
+---
+
+## Web versions
+
+Prefer a browser?
+
+- 🌐 [eucomplyhub.com/risk-class](https://eucomplyhub.com/risk-class) — interactive Tier 0 classifier (same logic as `quick_risk_class`)
+- 🔬 [eucomplyhub.com/annex3](https://eucomplyhub.com/annex3) — Tier 1 deep classifier (same as `classify_annex3`)
+- 🗺️ [eucomplyhub.com/crosswalk](https://eucomplyhub.com/crosswalk) — interactive EU AI Act ↔ ISO 42001 ↔ NIST mapping
+
+---
+
+## License
+
+MIT — see [LICENSE](LICENSE).
+
+## Contributing
+
+Issues + PRs welcome at https://github.com/eucomplyhub/mcp-eu-ai-act
+
+For questions about the audit methodology behind the tools, reach out: [piotr@eucomplyhub.com](mailto:piotr@eucomplyhub.com)
+
+---
+
+## Changelog
+
+### 0.1.0 — 2026-05-14
+
+- Initial release
+- Two tools: `quick_risk_class` + `classify_annex3`
+- 10+ frameworks covered
+- Claude Desktop / Cursor / Windsurf install instructions

package/dist/index.d.ts

@@ -0,0 +1,17 @@
+#!/usr/bin/env node
+/**
+ * @eucomplyhub/mcp-eu-ai-act
+ *
+ * MCP (Model Context Protocol) server exposing eucomplyhub.com's free EU AI Act
+ * compliance classifiers to AI assistants (Claude Desktop, Cursor, Windsurf, etc.)
+ *
+ * Two tools exposed:
+ *   1. classify_annex3 — Deep Annex III classification (Claude-powered)
+ *      Maps to: https://eucomplyhub.com/api/annex3-classify
+ *   2. quick_risk_class — 30-second multi-jurisdiction risk classifier
+ *      (deterministic, no LLM call — same logic as eucomplyhub.com/risk-class)
+ *
+ * @license MIT
+ * @see https://eucomplyhub.com
+ */
+export {};

package/dist/index.js

@@ -0,0 +1,331 @@
+#!/usr/bin/env node
+/**
+ * @eucomplyhub/mcp-eu-ai-act
+ *
+ * MCP (Model Context Protocol) server exposing eucomplyhub.com's free EU AI Act
+ * compliance classifiers to AI assistants (Claude Desktop, Cursor, Windsurf, etc.)
+ *
+ * Two tools exposed:
+ *   1. classify_annex3 — Deep Annex III classification (Claude-powered)
+ *      Maps to: https://eucomplyhub.com/api/annex3-classify
+ *   2. quick_risk_class — 30-second multi-jurisdiction risk classifier
+ *      (deterministic, no LLM call — same logic as eucomplyhub.com/risk-class)
+ *
+ * @license MIT
+ * @see https://eucomplyhub.com
+ */
+import { Server } from "@modelcontextprotocol/sdk/server/index.js";
+import { StdioServerTransport } from "@modelcontextprotocol/sdk/server/stdio.js";
+import { CallToolRequestSchema, ListToolsRequestSchema, } from "@modelcontextprotocol/sdk/types.js";
+const SERVER_NAME = "@eucomplyhub/mcp-eu-ai-act";
+const SERVER_VERSION = "0.1.0";
+const API_BASE = "https://eucomplyhub.com";
+// ─── Deterministic risk-class logic (mirror of /risk-class page) ─────────────
+const INDUSTRY_W = {
+    healthcare: [8, 10, 7, 9, 9],
+    publicSector: [9, 8, 9, 8, 9],
+    education: [7, 7, 8, 6, 6],
+    hr: [8, 8, 9, 6, 7],
+    retail: [3, 5, 4, 4, 4],
+    industrial: [4, 4, 3, 7, 4],
+    media: [4, 4, 6, 7, 4],
+    other: [5, 5, 5, 5, 5],
+};
+const ARCHETYPE_W = {
+    decisioning: [9, 5, 9, 8, 8],
+    generation: [4, 5, 5, 8, 4],
+    classification: [6, 5, 7, 6, 5],
+    recommendation: [5, 5, 7, 4, 5],
+    automation: [7, 5, 5, 9, 7],
+    forecasting: [4, 5, 6, 7, 4],
+};
+const IMPACT_W = {
+    internal: [2, 3, 3, 4, 2],
+    b2b: [4, 5, 5, 6, 5],
+    consumer: [6, 7, 7, 7, 7],
+    regulated: [10, 9, 10, 9, 10],
+};
+const AXES = ["decisionalImpact", "dataSensitivity", "biasRisk", "hallucinationCost", "regulatoryLoad"];
+function r1(x) {
+    return Math.round(x * 10) / 10;
+}
+function classifyRisk(industry, archetype, impact) {
+    const i = INDUSTRY_W[industry];
+    const a = ARCHETYPE_W[archetype];
+    const p = IMPACT_W[impact];
+    if (!i || !a || !p) {
+        throw new Error(`Invalid input. industry must be one of: ${Object.keys(INDUSTRY_W).join(", ")}. archetype: ${Object.keys(ARCHETYPE_W).join(", ")}. impact: ${Object.keys(IMPACT_W).join(", ")}.`);
+    }
+    const scores = {};
+    for (let k = 0; k < 5; k++) {
+        scores[AXES[k]] = r1((i[k] + a[k] + p[k]) / 3);
+    }
+    const m = Math.max(scores.decisionalImpact, scores.biasRisk, scores.regulatoryLoad);
+    let cls, summary;
+    if (m >= 9) {
+        cls = "Critical";
+        summary = "Re-scope before any build. Conformity assessment, fundamental-rights review and sector-specific approvals are mandatory before deployment.";
+    }
+    else if (m >= 7) {
+        cls = "High";
+        summary = "Engineer for conformity assessment, fairness evaluation and tamper-evident audit trail from day one. Annex III obligations apply.";
+    }
+    else if (m >= 5) {
+        cls = "Limited";
+        summary = "Ship with transparency disclosures, output grounding and a documented incident-response plan. Article 50 transparency applies.";
+    }
+    else {
+        cls = "Minimal";
+        summary = "Move fast — but maintain an audit trail and basic ISO/IEC 42001 hygiene. Re-classify when scope changes.";
+    }
+    const di = scores.decisionalImpact, br = scores.biasRisk, rl = scores.regulatoryLoad, hc = scores.hallucinationCost, ds = scores.dataSensitivity;
+    const frameworks = [
+        {
+            id: "eu_ai_act",
+            label: "EU AI Act",
+            region: "EU",
+            verdict: di >= 9
+                ? "Review for Art. 5 prohibited practices before any deployment."
+                : (di >= 7 || impact === "regulated")
+                    ? "High-risk system — full Annex III obligations apply (Articles 9–15, 26, 27)."
+                    : di >= 5
+                        ? "Limited risk — Article 50 transparency duties apply (user notification + content labelling)."
+                        : "Minimal risk — voluntary best-practice + ISO/IEC 42001 alignment.",
+        },
+        {
+            id: "nist_ai_rmf",
+            label: "NIST AI RMF",
+            region: "US / global",
+            verdict: br >= 8
+                ? "Govern + Measure priority — fairness, validity and reliability controls."
+                : hc >= 7
+                    ? "Measure + Manage — robustness and reliability emphasis."
+                    : "Map + Govern — full functional alignment across all four functions.",
+        },
+        {
+            id: "iso_42001",
+            label: "ISO/IEC 42001",
+            region: "Global",
+            verdict: rl >= 8
+                ? "Full AIMS implementation with sector annex controls (Clauses 4–10)."
+                : rl >= 6
+                    ? "Core AIMS implementation with documented risk treatment plan."
+                    : "Lightweight AIMS aligned with existing ISO 27001 program.",
+        },
+        {
+            id: "oecd",
+            label: "OECD AI Principles",
+            region: "50+ countries",
+            verdict: (br >= 7 || di >= 7)
+                ? "Human-centred values, transparency and accountability obligations."
+                : "Standard accountability + transparency posture sufficient.",
+        },
+        {
+            id: "singapore",
+            label: "Singapore Model AI Governance",
+            region: "Singapore",
+            verdict: (impact === "consumer" || impact === "regulated")
+                ? "Apply AI Verify testing toolkit before deployment."
+                : "Internal governance committee + risk impact assessment.",
+        },
+        {
+            id: "gdpr",
+            label: "GDPR · UK GDPR · DPDP · CCPA · LGPD · PIPEDA",
+            region: "Global privacy",
+            verdict: ds >= 7
+                ? "Strong: DPA(s), DPIA, DPO consultation, Article 22 / DPDP §11 automated-decision review."
+                : "Standard: lawful basis, transparency notices, data subject rights operationalised.",
+        },
+    ];
+    // PRC GenAI overlay
+    if (archetype === "generation") {
+        frameworks.push({
+            id: "prc_genai",
+            label: "PRC GenAI Interim Measures",
+            region: "China",
+            verdict: "Pre-launch security assessment + watermarking obligations apply for PRC users.",
+        });
+    }
+    // Sector overlays
+    if (industry === "healthcare") {
+        frameworks.push({
+            id: "hipaa_fda_samd",
+            label: "HIPAA / FDA SaMD / EU MDR-AI",
+            region: "Health",
+            verdict: "PHI minimisation, BAAs, and software-as-medical-device classification check required.",
+        });
+    }
+    else if (industry === "hr") {
+        frameworks.push({
+            id: "us_employment",
+            label: "EEOC + NYC AEDT + Colorado SB 24-205",
+            region: "US employment",
+            verdict: "Pre-deployment bias audit + candidate notice + annual disparate-impact testing.",
+        });
+    }
+    else if (industry === "publicSector") {
+        frameworks.push({
+            id: "us_public_sector",
+            label: "OMB M-24-10 + CoE AI Convention",
+            region: "Public sector",
+            verdict: "Rights-impact assessment + public AI use case inventory + human rights safeguards.",
+        });
+    }
+    return {
+        class: cls,
+        summary,
+        scores,
+        frameworks,
+        methodology: "Deterministic, rule-based classification. Same inputs always produce the same outputs. Logic adapted from @clustral/risk-compass (MIT). Triple-framework methodology by eucomplyhub.com.",
+        disclaimer: "This is a high-level orientation tool, not legal advice. For binding compliance mapping, consult an expert. See https://eucomplyhub.com/risk-class for the interactive web version.",
+    };
+}
+// ─── MCP Server setup ────────────────────────────────────────────────────────
+const server = new Server({
+    name: SERVER_NAME,
+    version: SERVER_VERSION,
+}, {
+    capabilities: {
+        tools: {},
+    },
+});
+// List available tools
+server.setRequestHandler(ListToolsRequestSchema, async () => {
+    return {
+        tools: [
+            {
+                name: "quick_risk_class",
+                description: "30-second deterministic multi-jurisdiction AI risk classification across EU AI Act, NIST AI RMF, ISO/IEC 42001, OECD, Singapore Model AI Governance, GDPR, and sector-specific overlays (HIPAA, NYC AEDT, Colorado SB 24-205). No LLM call. Same inputs always produce same outputs. Use this for quick orientation before deeper analysis.",
+                inputSchema: {
+                    type: "object",
+                    properties: {
+                        industry: {
+                            type: "string",
+                            enum: ["healthcare", "publicSector", "education", "hr", "retail", "industrial", "media", "other"],
+                            description: "Industry vertical. Use 'other' for fintech/banking/insurance.",
+                        },
+                        archetype: {
+                            type: "string",
+                            enum: ["decisioning", "generation", "classification", "recommendation", "automation", "forecasting"],
+                            description: "What the AI primarily does: decisioning (approve/deny, score, allocate), generation (draft, summarise), classification (label, route, triage), recommendation (rank, suggest), automation (multi-step actions), forecasting (predict outcomes).",
+                        },
+                        impact: {
+                            type: "string",
+                            enum: ["internal", "b2b", "consumer", "regulated"],
+                            description: "Who is affected: internal (employee-only), b2b (business customers), consumer (end users), regulated (healthcare, finance, public sector).",
+                        },
+                    },
+                    required: ["industry", "archetype", "impact"],
+                },
+            },
+            {
+                name: "classify_annex3",
+                description: "Deep Annex III classification using Claude (Anthropic) with full EU AI Act regulatory context. Maps your product against all 8 Annex III high-risk categories + Article 50 transparency + GPAI Article 53 + GPAI provider/deployer reasoning. Returns 5 priority remediation actions tailored to your stack. ~60 seconds (LLM-powered).",
+                inputSchema: {
+                    type: "object",
+                    properties: {
+                        company: {
+                            type: "string",
+                            description: "Company name being classified.",
+                        },
+                        industry: {
+                            type: "string",
+                            description: "Industry/vertical (e.g., 'HR-tech', 'fintech', 'healthtech', 'productivity', 'voice-AI').",
+                        },
+                        features: {
+                            type: "array",
+                            items: { type: "string" },
+                            description: "AI features in the product. Examples: 'content-generation', 'scoring', 'decision-making', 'api-consumer', 'own-model', 'recommendation', 'classification'.",
+                        },
+                        useCase: {
+                            type: "string",
+                            description: "Plain-English description of what the AI does and who it serves. Min 20 chars.",
+                        },
+                        euExposure: {
+                            type: "string",
+                            enum: ["eu-customers-output", "eu-employees-only", "no-eu", "considering-eu"],
+                            description: "EU exposure level. 'eu-customers-output' = serves EU customers with AI-influenced outputs. 'eu-employees-only' = internal tools used by EU employees only. 'no-eu' = no EU footprint. 'considering-eu' = planning EU expansion.",
+                        },
+                    },
+                    required: ["company", "industry", "features", "useCase", "euExposure"],
+                },
+            },
+        ],
+    };
+});
+// Handle tool calls
+server.setRequestHandler(CallToolRequestSchema, async (request) => {
+    const { name, arguments: args } = request.params;
+    if (!args) {
+        throw new Error("Missing arguments");
+    }
+    if (name === "quick_risk_class") {
+        const { industry, archetype, impact } = args;
+        try {
+            const result = classifyRisk(industry, archetype, impact);
+            return {
+                content: [
+                    {
+                        type: "text",
+                        text: JSON.stringify(result, null, 2),
+                    },
+                ],
+            };
+        }
+        catch (err) {
+            return {
+                content: [
+                    {
+                        type: "text",
+                        text: `Error: ${err.message}`,
+                    },
+                ],
+                isError: true,
+            };
+        }
+    }
+    if (name === "classify_annex3") {
+        try {
+            const response = await fetch(`${API_BASE}/api/annex3-classify`, {
+                method: "POST",
+                headers: { "Content-Type": "application/json" },
+                body: JSON.stringify(args),
+            });
+            if (!response.ok) {
+                const errText = await response.text();
+                throw new Error(`Classifier API returned ${response.status}: ${errText.slice(0, 200)}`);
+            }
+            const result = await response.json();
+            return {
+                content: [
+                    {
+                        type: "text",
+                        text: JSON.stringify(result, null, 2),
+                    },
+                ],
+            };
+        }
+        catch (err) {
+            return {
+                content: [
+                    {
+                        type: "text",
+                        text: `Error calling classifier: ${err.message}. Try the web version: https://eucomplyhub.com/annex3`,
+                    },
+                ],
+                isError: true,
+            };
+        }
+    }
+    throw new Error(`Unknown tool: ${name}`);
+});
+// Start server
+async function main() {
+    const transport = new StdioServerTransport();
+    await server.connect(transport);
+    console.error(`[${SERVER_NAME}] v${SERVER_VERSION} listening on stdio.`);
+}
+main().catch((err) => {
+    console.error("Fatal error:", err);
+    process.exit(1);
+});

package/package.json

@@ -0,0 +1,63 @@
+{
+  "name": "@eucomplyhub/mcp-eu-ai-act",
+  "version": "0.1.0",
+  "description": "MCP server for EU AI Act compliance — Annex III deep classifier + multi-jurisdiction risk classifier. Powers AI assistants (Claude Desktop, Cursor, Windsurf) to evaluate mid-market SaaS systems against EU AI Act, NIST AI RMF, ISO/IEC 42001, OECD, and more frameworks.",
+  "keywords": [
+    "mcp",
+    "model-context-protocol",
+    "eu-ai-act",
+    "annex-iii",
+    "ai-governance",
+    "compliance",
+    "audit",
+    "iso-42001",
+    "nist-ai-rmf",
+    "gdpr",
+    "ai-act-classifier"
+  ],
+  "author": {
+    "name": "Piotr Reder",
+    "email": "piotr@eucomplyhub.com",
+    "url": "https://eucomplyhub.com"
+  },
+  "license": "MIT",
+  "homepage": "https://eucomplyhub.com",
+  "repository": {
+    "type": "git",
+    "url": "https://github.com/eucomplyhub/mcp-eu-ai-act.git"
+  },
+  "bugs": {
+    "url": "https://github.com/eucomplyhub/mcp-eu-ai-act/issues"
+  },
+  "type": "module",
+  "main": "dist/index.js",
+  "bin": {
+    "mcp-eu-ai-act": "dist/index.js"
+  },
+  "files": [
+    "dist",
+    "README.md",
+    "LICENSE"
+  ],
+  "scripts": {
+    "build": "tsc",
+    "start": "node dist/index.js",
+    "dev": "tsx src/index.ts",
+    "prepublishOnly": "npm run build"
+  },
+  "dependencies": {
+    "@modelcontextprotocol/sdk": "^1.0.0"
+  },
+  "devDependencies": {
+    "@types/node": "^22.0.0",
+    "tsx": "^4.19.0",
+    "typescript": "^5.6.0"
+  },
+  "engines": {
+    "node": ">=18.0.0"
+  },
+  "publishConfig": {
+    "access": "public",
+    "registry": "https://registry.npmjs.org/"
+  }
+}