Python MCP server for WhatsApp. Consumes the Go bridge REST API and exposes tools to Claude.
13 tools with consistent naming (verb_noun pattern) and basic descriptions. All tools have input schemas with typed parameters and descriptions. However, descriptions are minimal (avg ~50 chars, well below the 194-char baseline), lack context on WHEN to use tools, and omit dependency hints. No output schemas documented. Error handling is absent, tools do not guide recovery or categorize failures. Missing idempotency guidance for write operations. No tool annotations (readOnlyHint/destructiveHint). Composition is reasonable (single responsibility per tool), but draft-then-confirm pattern for sends is underexplained.
Confirm and send a previously drafted WhatsApp message
Get details about a specific WhatsApp chat
List all WhatsApp chats
List messages from a WhatsApp chat
Mark a WhatsApp chat as read
Search WhatsApp contacts by name or phone number
Draft a WhatsApp audio/voice message (does not send immediately; requires confirm_send)
Draft a WhatsApp file/media message (does not send immediately; requires confirm_send)
Descriptions are minimal (avg ~50 chars vs 194-char baseline). Missing WHEN to use, prerequisites, and dependency hints. E.g., 'send_message' lacks explanation of draft-then-confirm workflow.
No output schemas documented. LLMs cannot infer what fields are returned or plan downstream calls. E.g., list_messages should document returned message structure (id, sender, timestamp, text).
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | C | 61 | <=2025-11-25 | v2 |
Draft a WhatsApp message (does not send immediately; requires confirm_send)
Draft a WhatsApp emoji reaction to a message (does not send immediately; requires confirm_send)
Draft a WhatsApp reply/quote message (does not send immediately; requires confirm_send)
Send a typing indicator to a WhatsApp chat
Set online/offline presence status in WhatsApp
No error handling guidance. Tools do not indicate what errors are retryable, user-fixable, or fatal. No recovery hints (e.g., 'Chat not found. Try list_chats() to find the correct JID.').
No tool annotations (readOnlyHint, destructiveHint, idempotentHint). LLMs cannot distinguish safe reads from destructive writes or understand retry safety. E.g., send_message should be marked destructiveHint=true.
Draft-then-confirm pattern (send_message → confirm_send) is underexplained. Descriptions do not clarify that send_message is non-destructive and requires confirm_send to finalize, or how draft_id links the two calls.