A message broker and adapter system for routing MCP tools across multiple CLI clients (Claude, Codex, Grok, Antigravity, Cursor, DCode, Desktop) to Telegram and other messaging platforms
C3 Broker exposes 4 tools with basic descriptions but critical gaps in schema completeness and parameter documentation. Tool names lack clear verb-noun structure (attach, reply, fetch_queue, stop_poll are vague about intent). Input schemas exist but lack proper type constraints and descriptions for parameters. The 'limit' parameter in fetch_queue accepts both string and number with minimal guidance. No output schemas documented. Error handling and recovery guidance absent. Descriptions are present but generic (10-100 chars), below the 50-200 char baseline for LLM optimization. No tool annotations (readOnlyHint, destructiveHint, idempotentHint) despite clear WRITE/READ_ONLY risk classifications in metadata.
Attach to a Telegram topic or channel for message routing
Retrieve queued messages from Telegram with configurable limit and all-or-nothing semantics
Send a reply message to an attached Telegram topic
Stop polling for new messages and close the current session (pull-only host; results must be retrieved via fetch_queue)
Tool names lack action verbs and clarity. 'attach' is ambiguous (attach to what? attach how?). 'fetch_queue' mixes verb and noun awkwardly. LLMs cannot infer intent from these names alone.
Input schemas lack proper type constraints and parameter descriptions. 'limit' parameter in fetch_queue accepts ['string', 'number'] with only a description mentioning '1-50 or all', no enum, no pattern, no validation guidance for LLM.
No output schemas documented. LLMs cannot plan downstream calls or extract required fields (e.g., message IDs, queue status) without knowing response structure.
Tool annotations missing despite clear risk classifications in metadata. WRITE tools (attach, reply, stop_poll) should declare destructiveHint=true; READ_ONLY tool (fetch_queue) should declare readOnlyHint=true. Agents cannot distinguish safe from unsafe operations.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 41 | 2026-07-28+ | v2 |
No error handling or recovery guidance. If attach fails (invalid topic), fetch_queue fails (queue empty), or reply fails (message too long), LLM receives no actionable next steps.