MCP (Model Context Protocol) server for WordPress AI chatbot plugin. Provides tools for managing conversations, searching knowledge base, retrieving site information, and sending messages to AI providers.
Server has 5 tools with complete schemas and descriptions. Naming follows verb_noun convention (search_, list_, get_, send_). Descriptions are present (avg ~120 chars) but lack actionable context for LLM selection. Parameters have types and descriptions, but lack constraints (enums, ranges, patterns). Output schemas are not documented. Error handling returns error objects but lacks recovery guidance. No tool annotations (readOnlyHint/destructiveHint). Security is reasonable (no secrets in params), but permission checks are not visible.
Get a conversation with all its messages. Returns the conversation details and message history.
Get WordPress site information and AI chatbot plugin configuration. Does not expose API keys.
List chat conversations with message counts. Returns recent conversations sorted by creation date.
Search the knowledge base and site content index. Returns matching articles, FAQs, and indexed pages.
Send a message to the AI chatbot and get a response. Uses the configured AI provider and includes knowledge base context.
Output schemas not documented. LLMs cannot plan downstream calls or extract required fields. get_conversation returns conversation + messages array, but structure is inferred from code, not declared in schema.
Parameters lack constraints. 'limit' in search_knowledge and list_conversations have no min/max bounds. LLMs may pass absurd values (limit=999999). 'query' in search_knowledge has no length or format guidance.
Error responses lack recovery guidance. execute() returns ['error' => message] but does not tell LLM what to do next. E.g., 'conversation_id is required' should suggest 'Call list_conversations() first to find a valid ID.'
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | C | 60 | 2026-07-28+ | v2 |
Tool annotations missing. send_message is destructive (WRITE risk) but has no destructiveHint. get_conversation is read-only but has no readOnlyHint. Agents cannot distinguish safe from unsafe tools.
Descriptions lack WHEN/WHY context. 'Search the knowledge base and site content index' does not explain when to call this vs. send_message. LLMs cannot distinguish between discovery and action tools.