A collaborative document and collection management system with MCP server integration for document operations, collection queries, and record management
Compendium provides 15 tools with consistent naming (verb_noun pattern: documents.*, collections.*, records.*) and basic descriptions. All tools have input schemas with type definitions and parameter descriptions. However, descriptions are terse (avg ~40 chars, well below the 194-char baseline), lack context on WHEN to use each tool, and omit guidance on state-modifying consequences. Output schemas are not documented. Error handling is minimal, no recovery guidance or actionable error messages. The tool set is well-composed (single responsibility per tool, proper chaining via IDs), but lacks the depth expected for production use.
Create a new collection
Delete a collection
List all collections in the workspace
Query a collection with optional filters
Create a new document
Delete a document
Get a specific document by ID
Descriptions are uniformly terse (30 - 50 chars) and lack context. 'Delete a document' does not explain consequences, when to use it vs. moveDocument, or whether deletion is reversible. LLMs cannot distinguish similar tools or plan multi-step workflows.
No output schemas documented. Callers cannot know what fields to expect from listDocuments, queryCollection, or getRecord. This forces LLMs to guess field names for downstream tool calls, increasing errors.
Destructive tools (deleteDocument, deleteCollection, deleteRecord) lack confirmation or dry-run support. Agents can permanently delete data without a safety gate.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 58 | 2026-07-28+ | v2 |
List all documents in the workspace
Move a document to a new parent or position
Create a new record in a collection
Delete a record
Get a specific record by ID
List records in a collection
Update a record's properties
Write a record with full property and view configuration
queryCollection filter parameter is complex (array of objects with propertyKey, op, value) but lacks examples or validation guidance in the description. LLMs may construct invalid filters.
No error handling guidance. Tools do not document what errors are retryable, what user-fixable errors look like, or how to recover from failures. Agents have no recovery path.