# eu-ai-act-compliance-mcp
> EU AI Act (Regulation (EU) 2024/1689) compliance checking, risk classification, and documentation generation. By MEOK AI Labs.

## Install
pip install eu-ai-act-compliance-mcp

## Auth
- Free tier: 10 calls/day, no API key needed
- Pro tier: unlimited, set MEOK_API_KEY env var
- All tools are read-only and stateless

## Tools

### quick_scan
Instant risk classification from a one-sentence AI system description. No API key required.
- `description` (str, required): One-sentence description of the AI system
- Returns: risk_level (prohibited/high-risk/limited-risk/minimal), matched_areas, top_3_obligations, deadline, penalty_range
- Use when: You need a fast risk assessment before deeper analysis

### deadline_check
All EU AI Act enforcement deadlines with days remaining. No parameters needed.
- Returns: List of milestone dates with days_remaining and applicable articles
- Use when: Checking which deadlines apply or how much time remains

### classify_ai_risk
Detailed risk classification against Article 5 (prohibited), Annex III (high-risk), Article 50 (limited-risk).
- `description` (str, required): Description of AI system, purpose, data, deployment context
- `caller` (str, optional): Identifier for rate limiting
- `api_key` (str, optional): MEOK API key for pro tier
- Returns: classification, confidence, prohibited_matches, high_risk_matches, limited_risk_triggers, analysis
- Use when: You need detailed Article-level classification with matched keywords

### check_compliance
42-point compliance audit against Articles 9-15 for high-risk AI.
- `system_name` (str, required), `purpose` (str, required), `data_types` (str, required), `decision_scope` (str, required)
- `has_risk_management`, `has_data_governance`, `has_technical_docs`, `has_logging`, `has_transparency_info`, `has_human_oversight`, `has_accuracy_testing` (all bool, default False): Current compliance posture for Articles 9-15
- `api_key` (str, optional): MEOK API key
- Returns: Per-article pass/fail checklist, compliance_score, gaps, remediation steps
- Use when: Running a structured compliance gap analysis

### generate_documentation
Generates Annex IV technical documentation template (Pro tier).
- Required: `system_name`, `provider_name`, `provider_contact`, `version`, `intended_purpose`, `description`, `data_description`, `architecture_description` (all str)
- Optional: `performance_metrics`, `risk_management_description`, `human_oversight_description` (str), `api_key` (str)
- Returns: Complete Annex IV markdown documentation template (8 sections)
- Use when: Generating technical documentation for conformity assessment

### assess_penalties
Calculate potential fines per Article 99.
- `violation_type` (str, required): "prohibited" | "high_risk_obligations" | "incorrect_information"
- `annual_global_turnover_eur` (float, optional): Annual global turnover in EUR
- `is_sme` (bool, default False): SME status for proportionate penalties
- `api_key` (str, optional): MEOK API key
- Returns: Penalty calculation, aggravating/mitigating factors
- Use when: Assessing financial risk exposure for non-compliance

### get_timeline
All implementation milestones from entry into force through full enforcement.
- `api_key` (str, optional): MEOK API key
- Returns: Chronological list of enforcement dates with applicable articles
- Use when: Planning a compliance roadmap

### audit_report
All-in-one comprehensive audit report combining classification, compliance, docs, and penalties (Pro tier).
- Takes all parameters from check_compliance + generate_documentation + assess_penalties
- Returns: Complete markdown audit report
- Use when: Compliance officers need a single comprehensive deliverable

### multi_jurisdiction_map
Map EU AI Act articles to equivalent UK, Singapore, Canada, US NIST requirements (Pro tier).
- `article` (str, required): EU AI Act article (e.g. "Article 5", "Article 9")
- `jurisdictions` (list, optional): Filter jurisdictions, default all
- `api_key` (str, optional): MEOK API key
- Returns: Cross-jurisdiction mapping of equivalent requirements
- Use when: Operating across multiple regulatory jurisdictions

### predict_risk_neural
Neural network risk prediction with 12 structured inputs (Pro tier).
- `system_name` (str, required): AI system name
- `uses_biometric` (bool), `uses_health_data` (bool), `uses_financial_data` (bool)
- `has_human_oversight` (bool), `affected_users` (int), `sector` (str)
- `has_documentation` (bool), `prior_incidents` (int), `deployed_cross_border` (bool)
- `model_explainable` (bool), `api_key` (str)
- Returns: Risk scores, violation probability, remediation urgency
- Use when: Quantitative risk scoring for portfolio-level AI governance

## Quick Example
```python
result = quick_scan(description="AI system that screens job applicants by analyzing CVs")
# Returns: risk_level="high-risk", matched Annex III Area 4 (Employment)
```
