A Craft CMS starter kit with MCP server integration for AI-ready content management, featuring Datastar for hypermedia interactions, SEO optimization, and AI bot tracking
This MCP server has severe definition quality issues. Only 2 tools are exposed, both with minimal schema documentation. The 'ping' tool lacks any visible input schema definition in the source code (PingController.php shows no explicit parameter registration). The 'api-catalog' tool returns an empty RFC 8288 linkset+json but has no documented output schema. Neither tool has parameter descriptions. Descriptions exist but are generic and do not follow LLM-optimized patterns (50-200 chars baseline). No error handling guidance, no input validation hints, and no composition support. Tool names lack action verbs ('api-catalog' is not verb_noun). The server appears to be a Craft CMS module wrapper around the stimmt/craft-mcp plugin, but tool definitions are not directly visible in the provided source, they are likely delegated to the plugin, making this a thin wrapper with inferred tool definitions.
Public endpoint that returns an empty API catalog in RFC 8288 linkset+json format for agent discoverability
Admin-only endpoint that returns system information including Craft version, environment, plugins, Datastar status, queue info, user details, PHP version, server info, and database driver
No input schemas visible for either tool. 'ping' and 'api-catalog' have no documented parameter types, constraints, or descriptions in the source code.
Tool names do not follow verb_noun convention. 'api-catalog' is a noun phrase, not an action. 'ping' is a verb but lacks clarity on what it returns and when to use it vs other discovery tools.
No parameter descriptions. Neither tool documents what inputs it accepts, what formats are valid, or what constraints apply. LLMs cannot infer parameter meaning from names alone.
Output schemas not documented. 'ping' returns 'system information' but no structured schema is provided. 'api-catalog' returns RFC 8288 linkset+json but the response structure is not formally defined for LLM parsing.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 31 | 2026-07-28+ | v2 |
Tool definitions appear inferred from plugin delegation (stimmt/craft-mcp) rather than directly visible in source.