MCP (Model Context Protocol) Server for axesCheck (check.axes4.com) PDF Accessibility Checker. Enables AI models to run official PDF/UA & WCAG compliance audits via stdio JSON-RPC.
Single tool with adequate naming (audit_pdf_axescheck) and clear description (194 chars, within baseline 34-392 range). Input schema present with proper type and description for pdf_path parameter. However, no output schema documented, LLMs cannot infer what the tool returns or plan downstream actions. No error handling guidance, no parameter validation constraints (e.g., file existence, size limits), and no recovery hints. Risk classification (READ_ONLY) is present but tool lacks idempotent/destructive annotations. Baseline: 90% of A+ tools have documented return types; this tool has none.
Submits a PDF document to axesCheck (check.axes4.com) for official PDF/UA (ISO 14289-1) and WCAG 2.1 A/AA machine-verifiable accessibility compliance evaluation according to the Matterhorn Protocol.
No output schema documented. LLMs cannot determine what fields the tool returns (e.g., compliance status, violations list, score, report URL). This blocks downstream tool chaining and forces the LLM to guess at response structure.
No error handling guidance. If the PDF is invalid, the file does not exist, or the axesCheck API is unreachable, the tool provides no recovery hints. LLMs cannot self-correct or suggest alternatives.
Input parameter lacks validation constraints. No mention of file size limits, supported PDF versions, or path format requirements. LLMs may pass invalid paths (e.g., relative paths, URLs, non-existent files) without guidance.
Tool description does not state what the tool returns or when to use it instead of alternatives. Baseline: 100% of A+ tools answer 'What does it do? When should the LLM call it? What does it return?'
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 49 | 2026-07-28+ | v2 |
STDIO transport only. Server is not remotely accessible and cannot be used by hosted MCP clients. This is a hard architectural limitation, not a quality issue, but it severely restricts deployment scenarios.