MCP server for accessing 1C:Enterprise platform context (types, methods, properties) from HBK documentation files
This Kotlin MCP server for 1С platform context search has fundamental quality issues. While all 7 tools are clearly defined with names, descriptions, and input schemas, the definitions suffer from: (1) incomplete or missing output schema documentation (critical for LLM planning), (2) descriptions written in Russian without English translations (blocks adoption in English-speaking agent ecosystems), (3) no parameter validation guidance or constraints beyond type declarations, (4) no error handling documentation or recovery guidance, (5) missing tool annotations (readOnlyHint, destructiveHint) despite all tools being READ_ONLY, (6) no documented parameter dependencies or prerequisites. The tools follow verb_noun naming conventions well (search, get_*) and parameters have type declarations, but lack the depth required for production-grade agent tooling. All tools are discovery/read-only, which reduces complexity, but the absence of output schemas and LLM-optimized descriptions significantly limits agent reasoning and error recovery.
Получает информацию о методе платформы по его названию
Получает информацию о свойстве платформы по его названию
Возвращает список конструкторов для указанного типа
Получает подробную информацию о типе платформы по его названию
Получает информацию о конкретном члене типа (методе или свойстве) по имени типа и имени члена
Возвращает список всех методов и свойств для указанного типа
Выполняет поиск элементов в контексте платформы 1С по текстовому запросу с опциональной фильтрацией по типу
All descriptions are in Russian, blocking use in English-speaking agent ecosystems. LLMs trained primarily on English documentation struggle with Russian descriptions during tool selection reasoning.
No output/return schema documented for any tool. LLMs cannot plan downstream calls or extract required fields (e.g., type_id, member_name) needed for chaining tools. Critical for discovery patterns where get_type_members result must feed into get_type_member.
No tool annotations (readOnlyHint, destructiveHint, idempotentHint) despite all tools being explicitly READ_ONLY. Agents cannot determine which tools are safe to retry or compose safely without explicit annotations.
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | F | 39 | <=2025-11-25 | v2 |
| 2026-03-09 | F | 24 | - | v1 |
Parameter 'limit' in search tool has no documented constraint bounds. Description says '(1-50, по умолчанию 10)' but no min/max validation rules documented. Agents may pass invalid values; error handling is undefined.
No error handling or recovery guidance. 7 custom exceptions defined (PlatformTypeNotFoundException, TypeMemberNotFoundException, etc.) but no documentation of when they occur or what the LLM should do (retry, ask user, try alternative tool).
Parameter 'type' in search tool (filter by type: method, property, type) has no enum constraint or explicit list of valid values. Description says 'type, method, property' but not as machine-parseable enum.
No documented parameter dependencies or prerequisites. Unclear whether 'get_type_member' requires calling 'get_type_members' first or if 'typeName' and 'memberName' can be discovered via search. Multi-step workflows not documented.