Loading ToolBench
An MCP server deployed on Cloudflare Workers that demonstrates OAuth 2.0 authorization flow with a simple arithmetic tool (add) and provides a web-based authorization interface.
Single tool 'add' has a trivial implementation with minimal real-world utility. Tool name lacks action verb (should be 'sum' or 'calculate_sum'). Description is extremely generic (10 chars: 'Adds two numbers together'), below the 20-char hard floor for scoring. Parameters 'a' and 'b' have minimal descriptions ('First number to add', 'Second number to add') that lack context on valid ranges, precision, or use cases. No output schema documented. No error handling guidance. This is a demo/stub tool, not production-grade.
Adds two numbers together
Tool name 'add' does not start with action verb. Should be 'sum', 'calculate_sum', or 'add_numbers' to clarify intent.
Tool description is 10 characters ('Adds two numbers together'), below 20-char minimum. Lacks context on when to use, valid ranges, precision, or return format.
No output schema documented. LLM cannot infer return type, precision, or error conditions. Should document: returns number, range, precision (int vs float).
Parameters 'a' and 'b' lack constraints (min/max, precision). Unbounded numbers invite absurd inputs (e.g., 1e308 + 1e308 = Infinity).
No error handling guidance. What happens on overflow, NaN, or type mismatch? LLM has no recovery path.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 28 | 2026-07-28+ | v2 |