How much of your Claude subscription allowance is left, and whether it will last to the reset.
Single tool 'check_usage' has a reasonable description (89 chars, within 10-1024 range) but CRITICAL deficiency: input schema is completely empty ({}). Tool definition is inferred from cli_test.go rather than explicit MCP registration in source. No parameter descriptions, no output schema documentation, no error handling guidance. Tool naming is acceptable (verb_noun pattern: 'check_usage') but cannot overcome schema absence.
Check Claude subscription usage and budget remaining across billing windows
Input schema is empty ({}). No parameters defined, no types, no descriptions.
No output schema documented. LLMs cannot plan downstream calls or extract fields from the response.
Tool definition inferred from test file (cli_test.go), not explicitly registered in MCP server code.
No error handling guidance. If the usage endpoint fails or returns unexpected data, LLM has no recovery path.
No parameter descriptions. Tool accepts no parameters, but if it did, LLMs would have no guidance on what each controls.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 34 | 2026-07-28+ | v2 |