Loading ToolBench
Provides read-only access to the EUDAMED public API: search and look up economic operators (actors), UDI device records, and reference/nomenclature data used by the EU medical device and IVD registration system.
Six tools with clear verb-noun naming (search_*, get_*) and comprehensive descriptions (avg 180 chars). All tools have proper JSON Schema with typed parameters and descriptions. Strong composition: search tools accept optional filters combined with AND, get tools fetch by ID. Pagination is mentioned in descriptions but not explicitly exposed as parameters. Output schemas are not documented in the source. Error handling is basic (no recovery guidance visible). No tool annotations (readOnlyHint, etc.) despite all being read-only operations.
Look up a single EUDAMED actor by its exact actor identifier (UUID).
Resolve a single reference data label by its numeric id, group code, and language.
Look up a single UDI device record by its exact Primary Device Identifier.
Search economic operators (manufacturers, authorised representatives, importers, competent authorities, etc.) registered in EUDAMED. All filters are optional and combined with AND; omit all of them to list actors. Automatically follows pagination and returns every match.
Search EUDAMED reference/nomenclature data: lookup tables used to resolve numeric ids to human-readable labels (e.g. risk classes, legislations, statuses, countries). All filters are optional.
Search UDI (Unique Device Identification) device records in EUDAMED by identifiers, names, manufacturer, risk classification, or legislation. All filters are optional and combined with AND. Automatically follows pagination and returns every match, with resolved human-readable labels for risk class, legislation, status, and market status.
Output schemas not documented. LLMs cannot plan downstream calls or extract fields without knowing response structure. search_actors, search_udi_devices, search_reference_data return lists but response format is opaque.
Pagination not exposed as parameters. Descriptions mention 'automatically follows pagination' but no page/limit/cursor parameters visible. LLMs cannot control result volume or navigate large datasets.
No tool annotations despite all tools being read-only. readOnlyHint should be applied to all six tools to signal safety to agents and prevent unnecessary confirmation prompts.
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 59 | <=2025-11-25 | v2 |
Error handling lacks recovery guidance. No visible error messages that tell LLMs what to do next (e.g., 'Try search_actors() with partial name'). Stack traces or bare error codes provide no actionable path forward.
search_udi_devices has 13 optional parameters with no guidance on mutual exclusivity or recommended combinations. LLMs may pass conflicting filters or miss that nomenclatureCode is client-side filtered (not server-side).