Loading ToolBench
A MCP server providing tools to retrieve employee counts by company name and residence counts by client name, with HTTP streaming transport
This server defines 2 tools with basic input schemas and descriptions, but falls well short of production standards. Both tools have names starting with action verbs (get_) and include descriptions, but the descriptions are minimal (33-40 chars), parameter documentation is sparse, output schemas are completely undocumented, and error handling is absent. The tools accept only a single string parameter each with minimal guidance on expected values or constraints. No pagination, no error recovery patterns, no security validation visible. The descriptions lack WHEN to use each tool, what it returns, or any prerequisites. This is a proof-of-concept implementation, not production-grade.
Récupère le nombre d'employés par nom société
Récupère le nombre de résidence par nom client
Descriptions are too short (33-40 characters) and lack critical information. Neither tool states what it returns, when to use it instead of alternatives, or any prerequisites. Baseline is 194 chars avg for production tools.
Output schemas are completely undocumented. LLMs cannot plan downstream operations or know what fields to extract. Production tools must document return types (100% of A+ tools do).
Single parameter per tool with no constraints, enums, or validation hints. 'companyName' and 'clientName' accept arbitrary strings with no guidance on format, length, or examples. No mention of what happens if the name is not found.
Inferred effective spec: 2026-07-28+.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | F | 48 | 2026-07-28+ | v2 |
| 2026-03-09 | F | 43 | - | v1 |
No error handling visible. If a company name or client name does not exist, the response is silent. Tools must return actionable errors: 'Company not found. Try search_companies() first.' (pattern:recovery-guide).
Parameter descriptions are minimal. 'Nom de la société' and 'Nom du client' are brief and do not explain expected format, case sensitivity, whether partial matches work, or what to do if the name is ambiguous. Baseline is 72 chars avg for production params.
No pagination or result limits documented. If there are multiple companies or residences with the same name, the tool response is unclear. Large result sets could exhaust context window without pagination.
Tool names are ambiguous: getEmployeeCountByCompanyName and getResidenceCountByClientName do not make it immediately clear that one returns an integer count. The names suggest they retrieve counts, which is good, but the output type is not documented.