Model Context Protocol (MCP) server for WeCom (WeChat Work) bot message sending
This server implements 7 WeCom message-sending tools with reasonable schema coverage but suffers from moderate gaps in parameter descriptions, lack of error handling guidance, and missing output schema documentation. All tools follow a clear send_* naming convention and include input schemas with type definitions. However, many parameter descriptions are minimal or generic, output structures are undocumented, and there is no evidence of error recovery patterns or operation classification (retryable vs. fatal). The tools are narrowly focused (good), but the definitions lack LLM-optimized guidance for when to use each variant.
Send an image (JPG/PNG, base64-encoded) through a WeCom bot webhook.
Send a Markdown message through a WeCom bot webhook. Supports headings, bold, links, quotes, etc.
Send a news message (article list) through a WeCom bot webhook. Accepts 1-8 articles.
Send a news notice template card with a cover image through a WeCom bot webhook.
Send a text message through a WeCom bot webhook. Supports @mentioning users by ID or mobile number.
Send a text notice template card through a WeCom bot webhook. Supports highlighted content, key-value pairs, and links.
No output schema documentation. Tools return results but provide no structured description of response fields (success status, message_id, timestamp, etc.). LLMs cannot determine what to extract or how to chain subsequent calls.
No error handling guidance. Descriptions do not explain what errors can occur, whether they are retryable, or what the agent should do (e.g., 'If webhook returns 401, credentials may be invalid'). Agents cannot reason about failure modes.
Minimal parameter descriptions for complex nested objects. send_text_notice_card and send_news_notice_card accept deeply nested structures (source.icon_url, horizontal_content_list[].keyname, card_action.url) but provide only one-sentence descriptions. LLMs need field-level guidance to construct valid payloads.
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | C | 64 | <=2025-11-25 | v2 |
Upload a file to the WeCom server (up to 20MB). Returns a media_id you can use to send file messages.
Missing parameter constraints. send_text (2048 byte limit), send_markdown (4096 bytes), send_image (2MB max, JPG/PNG only), and upload_file (20MB max) document limits in descriptions but not as formal min/max or enum constraints. LLMs may exceed limits without machine-readable validation.
No distinction between tools on when to use each. send_text, send_markdown, send_image, and send_news all send messages but the descriptions do not explain when to prefer one over another. An LLM might choose randomly between send_text and send_markdown for formatted content.
upload_file lacks context for downstream use. Description does not explain how to use the returned media_id in subsequent send_* calls. If an LLM needs to send a file, it must infer the connection to upload_file, upload, get media_id, then call send_* with that ID.
send_news articles array has 1-8 item constraint documented in description but not in schema minItems/maxItems. Machine validation is absent.