Loading ToolBench
Demo MCP server with OAuth 2.0 authentication running on Cloudflare Workers, featuring a remote authorization flow
Single tool 'add' with basic Zod schema and minimal description. Tool is properly registered in src/index.ts using McpServer.tool() with Zod validation, which is compliant. However, description lacks context, parameter descriptions are present but sparse, and no output schema is documented. The tool is READ_ONLY (risk classification noted) and functionally simple, but definition quality falls short of production standards for parameter hints, error handling guidance, and output structure documentation.
Adds two numbers together
Tool description is generic and lacks context about when/why to use it. 'Adds two numbers together' does not explain use cases, constraints, or return value structure.
Output schema is not documented. Callers must infer that the response is a text field with the sum. Documentation should explicitly state: 'Returns an object with a single text field containing the numeric sum as a string.'
No error handling guidance. No documentation of what happens if inputs are invalid (e.g., non-numeric). LLM has no recovery path if call fails.
Parameter descriptions ('First number to add', 'Second number to add') lack constraints. No mention of valid ranges, precision limits, or whether negative numbers are allowed.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | F | 7 | 2026-07-28+ | v2 |
| 2026-03-09 | D | 51 | - | v1 |
Single tool server limits composition. No discovery tools, no error context, no pagination patterns (though not needed here). Server is purely functional, not production-ready as a building block.