# ark-forge/mcp-eu-ai-act [Health: Active]

**Category:** ⚖️ Legal  
**Repository:** https://github.com/ark-forge/mcp-eu-ai-act  
**GitHub Stars:** 11  
**Views:** 3  
**Installs:** 0  
**Upvotes:** 0  
**Directory Page:** https://allmcps.com/mcp/ark-forge-mcp-eu-ai-act

## Description
EU AI Act compliance scanner that detects regulatory violations in AI codebases with risk classification and remediation guidance.

## Claude Desktop Quick Installation
Install path detected from listing signals. Uses `uvx` (confidence: high):

```json
"mcpServers": {
  "mcp-eu-ai-act": {
    "command": "uvx",
    "args": ["eu-ai-act-scanner"]
  }
}
```

## Documentation

## What ark-forge/mcp-eu-ai-act does

The ark-forge/mcp-eu-ai-act MCP server examines an AI project and produces compliance-oriented findings for the EU AI Act. It identifies AI framework usage in source code and configuration or manifest files, assigns or evaluates a risk category, and checks whether expected compliance documentation exists and contains relevant content.

The scanner supports 16 AI frameworks across Python, JavaScript, TypeScript, Go, Java, and Rust. Findings can be linked to specific EU AI Act articles, including checks associated with risk management, data governance, transparency, and technical documentation. The output is intended to help developers and legal or compliance teams identify missing work; it is not described as a substitute for professional legal advice.

## How it works

The MCP interface accepts a project path for most project-oriented operations. `scan_project` reports scanned files, AI-related files, detected frameworks, and model references. `check_compliance` evaluates document content on a 0–100 scale and returns pass or fail information with article mappings. Its documented default risk category is `limited`.

`generate_report` combines scanning and compliance checking into an executive summary and a technical breakdown. `suggest_risk_category` instead takes a plain-text description of an AI system and compares it with prohibited, high-risk, limited-risk, and minimal-risk criteria described in the README. `generate_compliance_roadmap` orders remediation tasks by deadline and estimated effort, while `generate_compliance_templates` returns starter Markdown documents for required records.

## Setup and configuration

The repository documents two setup paths. For the packaged CLI and MCP entry point, install the PyPI distribution:

```bash
pip install eu-ai-act-scanner
```

The documented MCP executable is `eu-ai-act-mcp`, which can be registered with Claude Desktop, Claude Code, or Cursor. The repository also provides a source setup using a virtual environment, `requirements.txt`, and `python3 server.py`.

An HTTP mode is available for CI/CD or remote clients by installing Uvicorn and starting `python3 server.py --http`; the example listens on port 8089. The free tier is described as requiring no sign-up or API key. Pro and Certified plans require an API key and unlock additional functionality.

## Tools and capabilities

The documented MCP tools include:

- `scan_project` for framework and model detection.
- `check_compliance` for content scoring and article mapping.
- `generate_compliance_roadmap` for deadline-aware remediation planning.
- `generate_report` for combined executive and technical output.
- `suggest_risk_category` for description-based classification.
- `generate_compliance_templates` for starter compliance documents.

The README states that the free tier exposes two tools, while Pro exposes ten. The excerpt documents six tools and notes additional v2 features, including Annex IV package generation and compliance-report certification.

## Limitations and notes

Free usage is limited to five scans per day. The feature table reserves full compliance checks, reports, roadmap generation, GDPR scanning, combined EU AI Act and GDPR analysis, and CI/CD integration for paid plans. The README references both August 2, 2026 and August 2, 2027 in different sections, so teams should verify the applicable deadline before relying on generated schedules.

Risk suggestions and automated document checks provide structured analysis, but the material does not establish that the output is a legal determination or complete audit evidence. Certified cryptographic proof is listed as a separate Trust Layer capability rather than a free MCP feature.

## Getting started with this ark-forge/mcp-eu-ai-act MCP server
Always refer to the official documentation for the most accurate and up-to-date information.

_Full upstream README: https://allmcps.com/mcp/ark-forge-mcp-eu-ai-act/readme_