Loading ToolBench
An MCP server deployed on Cloudflare Workers with OAuth 2.0 authentication, providing remote access to MCP tools through a web-based authorization flow.
Single tool 'add' is minimally defined. Tool name 'add' lacks a clear verb prefix, 'add_numbers' would be better aligned with the verb_noun pattern. Description is present but extremely brief (28 chars), falling below the 50-char minimum recommended for LLM clarity. Input schema is visible and properly typed (a: number, b: number) with parameter descriptions, but lacks any output schema documentation. No error handling guidance, no mention of what happens on edge cases (overflow, NaN, Infinity). The tool is READ_ONLY (marked correctly), but this is a trivial utility function, not a production-grade integration. Overall, this reads like a minimal demo rather than a tool designed for agent integration.
Adds two numbers together
Tool name 'add' lacks action-verb clarity. Should follow verb_noun pattern (e.g., 'add_numbers', 'compute_sum') to make intent obvious without reading description.
Description is 28 characters ('Adds two numbers together'), below the recommended 50 - 1024 char range. Too brief to convey when/why to use this tool vs alternatives. Lacks guidance on use cases, prerequisites, or constraints.
No documented output schema. LLMs cannot predict the structure of the response or plan downstream operations. Should explicitly document return type (e.g., { result: number }).
No error handling guidance. What happens if inputs are NaN, Infinity, or trigger numeric overflow? Should document constraints and recovery paths.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | F | 39 | 2026-07-28+ | v2 |
| 2026-03-09 | F | 49 | - | v1 |
Trivial tool with no practical agent use case. This is a math operation demo, not a production tool. Real MCP servers integrate with external services or complex business logic.