Loading ToolBench
Convert any documentation URL into a ready-to-run MCP server — 100% local, no API keys needed
Three tools with basic verb-noun naming (search_docs, get_source, list_sources) and short descriptions (18-35 chars). All tools have input schemas with typed parameters and descriptions. However, descriptions are below the 50-200 char baseline for LLM optimization, and output schemas are not documented. No error handling guidance, no parameter constraints (e.g., topK range), and no indication of pagination support for list_sources. Tools are read-only (good security posture) but lack composition hints and chaining IDs.
Get all chunks from a specific source URL
List all indexed documentation sources
Search indexed documentation by semantic query
Descriptions are 18-35 characters, well below the 50-200 char baseline for LLM optimization. 'Search indexed documentation by semantic query' lacks WHEN to use it, prerequisites, or expected output structure.
Output schemas are not documented. LLMs cannot plan downstream calls or extract chaining IDs (e.g., source URLs from list_sources to pass to get_source). Responses are opaque.
search_docs topK parameter lacks range constraints. No min/max specified; LLMs could pass 0, negative, or absurdly large values. Should constrain to e.g. 1-100.
No error handling guidance. If search_docs returns no results, or get_source fails on invalid URL, LLMs have no recovery path. Errors should suggest alternatives (e.g., 'Try a broader query' or 'Available sources: ...').
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 49 | 2026-07-28+ | v2 |
list_sources has no pagination support visible. If documentation grows large, returning all sources at once could exhaust context. Should support limit/offset and return total count.