MCP (Model Context Protocol) server for Claudexia — query your balance, usage, and pricing from any MCP-compatible client (Claude Desktop, Cursor, Claude Code).
Three tools with clear verb-noun naming (get_*, list_*) and basic descriptions. However, critical gaps undermine quality: parameter descriptions are missing entirely (days param in claudexia_get_usage has no description in the schema), output schemas are undocumented, and error handling provides no recovery guidance. The claudexia_list_models tool returns hardcoded static data with a comment acknowledging the API is unavailable, indicating incomplete implementation. No tool annotations (readOnlyHint, etc.) despite all being read-only operations. Descriptions are adequate (50-150 chars) but lack actionable context for LLM selection.
Get the current Claudexia account balance in USD. Use this when the user asks 'how much do I have left' or 'what's my balance'.
Get Claudexia usage stats for the last N days (default 7). Returns total tokens and cost.
List models available on Claudexia with their per-million-token pricing.
Parameter descriptions missing from input schemas. The 'days' parameter in claudexia_get_usage has no description in the JSON schema, violating pattern:tool-description. LLMs cannot infer that 'days' accepts 1-90 or defaults to 7.
Output schemas completely undocumented. No tool declares what fields it returns (e.g., balance_usd, total_tokens, total_cost_usd). LLMs cannot plan downstream calls or extract structured data without knowing response shape.
Error handling provides no recovery guidance. Errors return raw messages like 'Claudexia API 401: ...' without actionable next steps. Pattern:recovery-guide requires errors to tell the LLM what to do (e.g., 'Check CLAUDEXIA_API_KEY environment variable').
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 52 | 2026-07-28+ | v2 |
claudexia_list_models returns hardcoded static data with a TODO comment ('replace with real fetch when api.claudexia.tech exposes /v1/models'). This is incomplete implementation, not production-ready.
No tool annotations despite all tools being read-only. Missing readOnlyHint in tool definitions prevents clients from optimizing caching and retry logic. Current spec (2026-07-28) supports tool annotations.