Generate a lean MCP server from any OpenAPI spec
This is a code generator that produces MCP servers from OpenAPI specs. The 32 example tools (Notion + Petstore) show consistent structural patterns but critical quality gaps. Tool names use verb_noun format (retrieve_ablock, update_pet) which is good. However, descriptions are extremely terse (5-20 chars: 'Retrieve a block', 'Update a pet'), far below the 50-200 char baseline for LLM-optimized descriptions. Parameter descriptions are minimal ('id', 'page_size') with no context on format, constraints, or purpose. Input schemas are present but sparse, most params lack descriptions entirely. No output schemas documented. Error handling is absent from visible code. Tool 11 (update_page_properties) has NO visible input schema at all. The generator appears functional for basic OpenAPI→MCP conversion but produces low-quality tool definitions unsuitable for production agent use.
Add a pet
Append block children
Create user
Create users with list input
Delete a block
Delete order
Delete pet
Delete user
Descriptions are critically short (5-20 chars). LLM-optimized descriptions should be 50-200 chars explaining WHAT, WHEN, and any prerequisites. Current descriptions ('Retrieve a block', 'Update a pet') provide no context for tool selection.
Parameter descriptions are missing or trivial ('id', 'page_size'). Every parameter must explain its purpose, format, constraints, and valid values. Current params lack this context entirely.
Output schemas are not documented. LLMs need to know what fields to expect in responses to plan downstream tool calls and extract data correctly. No visible return type documentation.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 34 | 2026-07-28+ | v2 |
Find pets by status
Find pets by tags
Get inventory
Get order by ID
Get pet by ID
Get user by name
Login user
Logout user
Place order
Query a database
Retrieve a block
Retrieve a database
Retrieve a Page
Retrieve a Page Property Item
Retrieve a user
Retrieve block children
Retrieve comments
Update a block
Update a database
Update page properties
Update a pet
Update pet with form
Update user
Upload file
No error handling or recovery guidance. Tools provide no actionable error messages. LLMs cannot determine if errors are retryable, user-fixable, or fatal.
No tool annotations (readOnlyHint, destructiveHint, idempotentHint). Risk levels are assigned in the evaluation but not exposed to the MCP client. Agents cannot distinguish safe reads from destructive operations.