Loading ToolBench
Remote Model Context Protocol server for Craft document management, authenticated via GitHub OAuth
Server has 8 tools with mostly complete schemas and descriptions, but significant gaps reduce quality. 4 tools are archived/non-functional (deleteText, getCollectionItems, createCollectionItems, updateCollectionItems), leaving only 4 active tools. Active tools have good parameter descriptions and enum constraints (e.g., insertText variant parameter), but descriptions are inconsistent in length and depth. listDocuments lacks any description. Error handling is present but generic. No tool annotations (readOnlyHint, destructiveHint, idempotentHint) despite clear risk levels. Output schemas are not documented. Composition is reasonable, tools chain well (readDocument → insertText/updateText), but archived tools suggest incomplete feature parity.
Add new items to a collection
Delete a page or heading and all its content
Browse and read collection items with filtering
Insert markdown content into the document
Show available documents
Read document structure and content with flexible depth control
Update existing collection items
listDocuments has no description, LLM cannot determine when to call it or what it returns
4 of 8 tools are archived and non-functional (deleteText, getCollectionItems, createCollectionItems, updateCollectionItems), they appear in the tool list but are disabled, creating confusion about server capabilities
No tool annotations (readOnlyHint, destructiveHint, idempotentHint) despite clear risk levels (READ_ONLY, WRITE, DESTRUCTIVE), LLM cannot reason about side effects or retry safety
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 53 | 2026-07-28+ | v2 |
Update the markdown content of a specific block
Output schemas are not documented for any tool, LLM cannot plan downstream calls or extract required fields
Error messages are generic ('Failed to...') without recovery guidance, LLM cannot self-correct or determine if retry is safe