A remote Model Context Protocol (MCP) server using Hono web framework that exposes messaging tools (e.g., Telegram) over HTTP. Each request is scoped to a specific bot token, allowing multi-tenant usage. Authentication is handled via Clerk OAuth.
Single tool 'telegram' has a clear action verb and reasonable description (54 chars). Input schema is properly defined with two parameters (chatId, message), both with descriptions. However, the description is below the 10-1024 char baseline sweet spot (194 chars avg for A+ tools), and lacks critical context about prerequisites (bot token setup), error conditions, and when to use it. No output schema is documented, the code shows it returns {content, structuredContent} but this is not formally declared. No error handling guidance. Tool lacks idempotent/destructive hints. Overall definition is functional but sparse.
Send a Telegram message.
Output schema not documented. Code returns {content, structuredContent} but LLMs cannot infer the structure. Agents need to know what fields to expect for downstream planning.
Tool description lacks critical context: no mention of prerequisites (Telegram bot token required), error conditions (invalid chatId, API failures), or when to use it vs alternatives. Baseline for A+ tools is 194 chars; this is 54 chars.
No error handling guidance. If Telegram API returns 'chat not found' or 'bot blocked', the LLM has no recovery path. Should document retryable vs fatal errors and suggest next steps.
Missing tool annotations. Tool is destructive (sends a message) but lacks destructiveHint annotation. Agents cannot reason about side effects or plan rollback strategies.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 59 | 2026-07-28+ | v2 |
Parameter descriptions are minimal. 'Telegram chat ID' and 'Message text to send' lack format guidance (e.g., is chatId numeric or string? Max message length?). Baseline for A+ params is 72 chars; these are ~20 chars each.