The EU AI Act is now in force. Non-compliant AI systems face fines up to 35M or 7% of global revenue. This MCP server scans your codebase, detects AI frameworks, classifies risk level, and tells you exactly what's missing.
3 commands. 2 minutes. Know where you stand.
git clone https://github.com/ark-forge/mcp-eu-ai-act.git
cd mcp-eu-ai-act
python3 example_usage.pyLearn more: arkforge.fr/mcp-eu-ai-act
| Step | Action | Output |
|---|---|---|
| 1 | Scan your project | Detects OpenAI, Anthropic, HuggingFace, TensorFlow, PyTorch, LangChain |
| 2 | Classify risk level | Unacceptable / High / Limited / Minimal (per EU AI Act) |
| 3 | Check compliance | Pass/Fail on each regulatory requirement |
| 4 | Report findings | JSON report with actionable recommendations |
smithery install @arkforge/mcp-eu-ai-actSmithery is the official MCP server registry.
git clone https://github.com/ark-forge/mcp-eu-ai-act.git
cd mcp-eu-ai-act
python3 server.pyNo dependencies required. Pure Python 3.8+.
from server import MCPServer
server = MCPServer()
# Scan a project for AI frameworks
scan = server.handle_request("scan_project", {"project_path": "/your/project"})
# Check compliance (limited risk by default)
check = server.handle_request("check_compliance", {
"project_path": "/your/project",
"risk_category": "high" # or "limited", "minimal", "unacceptable"
})
# Generate a full compliance report
report = server.handle_request("generate_report", {"project_path": "/your/project"})See examples/ for complete runnable examples.
Scans a project to detect AI model usage.
// Input
{ "project_path": "/path/to/project" }
// Output
{
"files_scanned": 150,
"ai_files": [{"file": "src/main.py", "frameworks": ["openai", "langchain"]}],
"detected_models": {"openai": ["src/main.py"], "langchain": ["src/main.py"]}
}Verifies EU AI Act compliance for a given risk category.
// Input
{ "project_path": "/path/to/project", "risk_category": "limited" }
// Output
{
"risk_category": "limited",
"compliance_status": {"transparency": true, "user_information": true, "content_labeling": false},
"compliance_score": "2/3",
"compliance_percentage": 66.7
}Generates a complete compliance report with recommendations.
// Input
{ "project_path": "/path/to/project", "risk_category": "high" }
// Output
{
"report_date": "2026-02-09T10:30:00",
"scan_summary": {"files_scanned": 150, "ai_files_detected": 5},
"compliance_summary": {"risk_category": "high", "compliance_percentage": 50.0},
"recommendations": ["Missing: Risk management system", "Missing: Human oversight documentation"]
}| Category | Examples | Key Requirements | Max Fine |
|---|---|---|---|
| Unacceptable | Social scoring, mass biometric surveillance | Prohibited | 35M / 7% revenue |
| High | Recruitment, credit scoring, law enforcement | Full documentation, risk management, human oversight | 15M / 3% revenue |
| Limited | Chatbots, recommendation systems, content generation | Transparency, user information, content labeling | 7.5M / 1.5% revenue |
| Minimal | Spam filters, video games | No specific obligations | - |
- OpenAI - GPT-3.5, GPT-4, OpenAI API
- Anthropic - Claude, Anthropic API
- HuggingFace - Transformers, pipelines, models
- TensorFlow - Keras, .h5 models
- PyTorch - .pt, .pth models
- LangChain - LLM chains, agents
Works with any MCP-compatible client:
- Claude Code
- VS Code with MCP extension
- CI/CD pipelines (GitHub Actions, GitLab CI)
- Custom integrations via JSON protocol
- Integration with EU compliance databases
- Multi-language support (FR, EN, DE, ES)
- PDF report export
- CI/CD native integration (GitHub Actions, GitLab CI)
- Automatic compliance documentation generation
MIT License - See LICENSE for details.
Built by ArkForge | Documentation