MCP server that provides unified access to organizational knowledge across multiple platforms (local docs, Guru, Notion)
This server defines 9 tools across Guru and Notion integrations. While all tools have descriptions and visible schemas in the source code, quality varies significantly. Tool naming follows verb_noun conventions adequately (guru-list-cards, notion-get-page, etc.), but parameter descriptions are inconsistent and often generic. Input schemas are present with types, but many parameters lack actionable constraint guidance. Output schemas are undocumented. Error handling is minimal, no recovery guidance, categorization, or actionable error messages visible in the source. The server falls in the 'Fair' range: definitions exist but lack the precision and error-handling maturity expected of production tools.
List and optionally download attachments from a Guru card
List and search Guru cards with optional filters
Read the full content of a specific Guru card by ID
A simple hello world example tool
Create a new page in the Notion database from markdown content
Export a Notion page as markdown
Get the full content of a specific Notion page by ID
Output schemas undocumented. Tool descriptions state WHAT is done but not what structure is returned. LLMs cannot plan downstream calls or extract required fields without documented output shape.
Error handling absent. No recovery guidance, categorization of retryable vs fatal errors, or actionable error messages visible. Tools fail opaquely with no guidance for agent self-correction.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | D | 50 | 2026-07-28+ | v2 |
| 2026-03-09 | F | 48 | - | v1 |
List, query, and search pages in the Notion database with optional filtering and sorting
Update an existing Notion page with new markdown content
Parameter descriptions generic and lack actionable constraints. 'Maximum number of results (max: 50)' buried in text rather than as formal min/max constraints. 'Field to sort by' with no enum of valid fields. LLMs cannot infer constraints from descriptive text alone.
Destructive operations (notion-create-page, notion-update-page) have no confirmation or dry-run support. Agents can make mistakes without a chance to review before commit.
No tool annotations. Tools declaring READ_ONLY and WRITE risk levels but MCP server does not expose readOnlyHint/destructiveHint/idempotentHint metadata for LLM safety reasoning.
Parameter descriptions inconsistent in depth. 'Search terms for card title/content' vs 'Whether to download the first attachment and return its content'. Variation signals incomplete review.