Loading ToolBench
Model Context Protocol server for the French Government's API Recherche d'Entreprises. Provides comprehensive business search capabilities including text-based search, geographic filtering, activity classification, certifications, financial data, and person-based queries.
Two tools with comprehensive parameter schemas and detailed descriptions. search_businesses is exceptionally well-documented with 47 distinct parameters, each with clear descriptions explaining constraints, formats, and examples. search_businesses_nearby is similarly structured with 10 focused parameters. Both tools follow action-verb naming conventions and include usage examples. However, the server lacks output schema documentation, critical for agents to understand what fields to expect in responses. No error handling guidance is visible in tool definitions. Tool descriptions are lengthy (1000+ chars) and could be more concise per token-optimization baselines (ideal: 50-200 chars). No tool annotations (readOnlyHint, destructiveHint, idempotentHint) despite READ_ONLY risk classification. Parameter descriptions are thorough but inconsistent in format, some reference external standards (INSEE, SIREN) without clear guidance on where agents find valid values.
Search for French businesses, associations, and public services using text-based criteria. This comprehensive search tool allows filtering by business name, address, directors, elected officials, activity codes, certifications, and many other criteria. **Key Features:** - Search by business name, address, or person names - Filter by business activity (NAF/APE codes) - Geographic filtering (postal codes, departments, regions) - Filter by business size, legal status, certifications - Financial data filtering (revenue, net result) - Support for associations, individual entrepreneurs, public services **Usage Examples:** - Search by name: {"q": "la poste"} - Search by activity: {"activite_principale": "62.01Z"} - Search in Paris: {"code_postal": "75001,75002"} - Search associations: {"est_association": true} - Search bio-certified businesses: {"est_bio": true}
Search for French businesses, associations, and public services near specific geographic coordinates. This tool enables geographic proximity-based discovery of organizations within a specified radius, with optional filtering by business activity and certifications.
No output schema documentation. Tools define inputs comprehensively but never specify what fields/structure responses contain. Agents cannot plan downstream operations or extract required data without trial-and-error.
Tool descriptions are verbose (1000+ characters) when best practice is 50-200 chars. Excessive length wastes tokens and buries key decision points for LLM tool selection. Example: search_businesses description includes full usage examples that should be in documentation, not in-context.
No tool annotations despite READ_ONLY classification. search_businesses and search_businesses_nearby are marked READ_ONLY in metadata but lack readOnlyHint in schema. Missing destructiveHint/idempotentHint annotations prevent agents from reasoning about safe retry/composition patterns.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | D | 58 | 2026-07-28+ | v2 |
| 2026-03-09 | F | 49 | - | v1 |
No error handling guidance in tool definitions. No description of what happens on invalid input (malformed coordinates, out-of-range radius), missing required parameters, or API failures. Agents receive raw errors with no recovery hints.
Parameter constraints reference external standards (INSEE, SIREN, NAF/APE codes) without guidance on discovery. Agents cannot know valid activity codes or legal nature values without external lookup. Consider providing reference URLs or suggesting a discovery tool.
Pagination parameters (page, per_page) lack documented defaults and max limits. Description says 'Maximum is 25' but doesn't clarify if that's a hard API limit or server-side enforcement. Unclear defaults invite agents to pass invalid values.