MCP server that classifies AI systems under the EU AI Act (Regulation (EU) 2024/1689) — risk tier, obligations and deadlines — for AI assistants like Claude and Cursor. By Disclos.
Four well-defined tools with clear, domain-specific descriptions (avg 180 chars). All tools have complete input schemas with typed parameters and descriptions. Tool names follow verb_noun convention (classify_, explain_, framework_). Schemas use enums for constrained inputs (tier parameter). However, output schemas are not formally documented, responses are rendered as plain text rather than structured objects. Error handling is minimal (only 'Unknown tool' and 'Unknown tier' cases visible). No tool annotations (readOnlyHint, idempotentHint) despite all tools being read-only. Parameter descriptions are good but could be more concise (some exceed 150 chars). Missing pagination, rate limiting, and recovery guidance in error messages.
Classify an AI system or product under the EU AI Act (Regulation (EU) 2024/1689) and return its risk tier, obligations and deadlines. Map the user's description to the boolean flags below. When unsure, ask the user or choose the more cautious answer. Returns a structured 'scope finding'.
Return the EU AI Act enforcement timeline — the key dates and what takes effect on each.
Explain a single EU AI Act risk tier in detail: what it means, the obligations, and the deadline.
Map an EU AI Act risk tier to the equivalent controls in adjacent governance frameworks — ISO/IEC 42001, NIST AI RMF, and GDPR. Use when the user asks how the EU AI Act lines up with ISO 42001, NIST, GDPR, or their existing compliance program.
Output schemas not documented. Tools return plain-text responses (renderFinding, renderTimeline, etc.) rather than structured JSON objects. LLMs cannot parse or chain results reliably.
No tool annotations. All four tools are read-only (no side effects), but readOnlyHint is not set in tool definitions. Agents cannot distinguish safe tools from destructive ones.
Error handling lacks recovery guidance. 'Unknown tier' error does not suggest valid options or next steps. No actionable error messages to guide LLM self-correction.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | C | 63 | 2026-07-28+ | v2 |
Parameter descriptions are verbose (some 150+ chars). Baseline for A+ tools is 72 chars avg. Reduces token efficiency and buries key constraints.
No pagination or result limits documented. eu_ai_act_timeline and framework_crosswalk return unbounded lists. Should cap results and document limits in descriptions.