A Next.js application that integrates MongoDB with ChatGPT via MCP (Model Context Protocol) to provide todo list management capabilities with interactive widgets
7 tools with clear verb-noun naming and basic schemas. All tools have descriptions (avg 60 chars, below 194-char baseline). Input schemas present with type definitions and parameter descriptions. However, output schemas are not documented in the source code, and descriptions lack depth about when/why to use each tool or what they return. No error handling guidance visible. Tool annotations (destructiveHint, readOnlyHint) are declared but not shown in schema definitions. Composition is sound, tools are single-responsibility and chainable via IDs.
Add a new item to a todo list
Mark a todo item as completed
Create a new todo list with a title
Delete a todo item from a list
Retrieve a todo list by ID
Fetch all todo lists with their items
Create or update multiple todo items in a list
Output schemas not documented. LLMs cannot infer what fields each tool returns, forcing them to guess downstream field names and risk broken tool chains.
Descriptions are brief (avg 60 chars vs 194-char baseline) and lack context about when to use each tool, prerequisites, or what data is returned. E.g., 'Create a new todo list with a title' does not explain what ID is returned or how to use it in subsequent calls.
No error handling guidance. If a listId is invalid or an item does not exist, the LLM receives no recovery hint. Should return actionable errors like 'List not found. Call list_todo_lists() to see available lists.'
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | C | 60 | 2025-06-18+ | v2 |
upsert_todo_items accepts an array of items with optional 'id' field but does not document per-item success/failure. If one item fails, does the entire operation fail or return partial results? Ambiguous behavior.
list_todo_lists has no pagination parameters (limit, offset, cursor). If there are hundreds of lists, returning all of them wastes tokens and risks context window exhaustion.