Loading ToolBench
A Model Context Protocol server that provides tools for searching, listing, and retrieving MDX/Markdown documentation content. Serves documentation via HTTP transport with session management.
The server implements 3 read-only tools for documentation search and retrieval via HTTP transport. All tools have names starting with action verbs (search-, list-, get-) and basic descriptions present. However, definitions lack depth: descriptions are minimal (12-41 chars, below baseline 194), parameter descriptions are generic, output schemas are completely undocumented, and there is no error handling guidance. Tool names are clear but descriptions do not explain WHEN to use each tool vs alternatives. Parameters lack constraints (e.g., list-docs 'limit' accepts any number, no min/max documented). No pagination guidance despite list-docs returning potentially large result sets. The server is functional but falls short of production-grade tool quality, descriptions are too terse to guide LLM selection confidently, and lack of output schema documentation forces LLMs to guess at response structure.
get document markdown content by URI
list documents by category
search for documentation by query
Tool descriptions are critically short (12-41 chars vs baseline 194 chars). Descriptions lack WHEN/WHY context, making LLM selection difficult. Example: 'search for documentation by query' does not explain search scope, result ranking, or relationship to list-docs.
Output schemas are completely undocumented. LLMs cannot predict what fields will be returned or how to chain results to downstream tools. No documentation of list-docs pagination behavior, result structure, or total count.
list-docs 'limit' parameter has no min/max constraints. LLM could pass absurd values (0, -1, 999999) causing performance issues. No guidance on default behavior when limit=0 (all results? none?).
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | F | 10 | <=2025-11-25 | v2 |
| 2026-03-09 | F | 38 | - | v1 |
No pagination guidance or next_cursor/offset support in schema despite list-docs returning potentially large doc sets. Returning unbounded lists risks context window exhaustion.
Parameter descriptions are generic and missing implementation details. Example: 'document category' for list-docs category param, does not state if category is required, what valid values are, or how to discover available categories.
No error handling or recovery guidance in tool definitions. Tools provide no insight into what errors are retryable, what prerequisites must be met, or how to handle missing docs gracefully.
list-docs sortBy enum values ('path', 'name') are not described. LLM does not know what 'path' sorting means (alphabetical by URI? by directory depth?) or when to prefer one over the other.
get-doc-content returns raw markdown content but schema does not specify format (plain string? with metadata? structured fields?). No guidance on frontmatter handling visible in tool definition.