MCP server for Knowtation Hub — personal knowledge and content system with multi-agent support, vault management, proposals, and hosted canister integration
28 tools with basic schemas and descriptions, but significant gaps in parameter documentation and output schema clarity. Most tools have 1-5 parameters with type info, but descriptions are often generic (10-50 chars). No visible error handling guidance, no output schemas documented, no tool annotations (readOnlyHint/destructiveHint). Parameter descriptions lack constraints (enums, ranges, formats). Many tools conflate multiple concerns (e.g., 'enrich' is vague; 'hub_create_proposal' mixes proposal creation with change specification). WRITE tools lack confirmation/dry-run patterns. Naming is mostly verb-noun but some tools are ambiguous (e.g., 'enrich', 'relate'). No evidence of idempotency guarantees or recovery guidance in error cases.
Retrieve a specific attachment
List attachments in the vault
Find notes that link to a given note
Capture content into the vault inbox
Cluster notes semantically using k-means on embeddings
Enrich a note with additional context or metadata
Export vault notes to various formats
No output schemas documented for any tool. LLMs cannot plan downstream calls or extract required fields (e.g., does search() return note_id, path, or both?). Violates pattern:tool and pattern:response-shaper.
Parameter descriptions lack constraints. 'limit' appears in 10+ tools with no min/max (e.g., search, list_notes, cluster). 'format' in export has no enum. LLMs will pass invalid values (limit=999999, format='pdf'). Violates pattern:constrained-input.
WRITE tools (write, capture, transcribe, vault_sync, import, import_url) lack confirmation/dry-run patterns and no error recovery guidance. Agents cannot safely retry or undo. Violates pattern:confirmation-request and pattern:recovery-guide.
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 56 | <=2025-11-25 | v2 |
Extract checkbox tasks from notes with optional filtering
Build a document tree structure for a note showing parent-child relationships
Get metadata facets (tags, projects, dates) from vault notes
Retrieve a single note by path from the vault
Get the hierarchical outline structure of a note
Get the source markdown for a specific section within a note
Create a proposal for vault changes in hosted Hub
Import notes from various sources
Import content from a URL
Index vault notes for semantic search
List notes in the vault with optional filtering
Propose attaching media to a note
Propose adding an external media link
List media import consent records
Find semantically related notes to a given note
Search vault notes using semantic or keyword search
Generate a summary of a note using LLM
Suggest tags for a note based on semantic similarity
Transcribe audio content using Whisper
Synchronize vault with git repository
Write or update a note in the vault
Ambiguous tool names: 'relate' (find related notes?), 'enrich' (add metadata? generate summary?), 'cluster' (group notes? return cluster assignments?). LLMs cannot infer intent from names alone. Violates pattern:tool naming rules.
No tool annotations (readOnlyHint, destructiveHint, idempotentHint). Agents cannot distinguish safe reads from destructive writes. 'write' and 'vault_sync' should be marked destructiveHint=true. Violates current MCP spec (2026-07-28).