A remote MCP server with OAuth 2.0 authentication provider, demonstrating authorization flows and MCP tool registration on Cloudflare Workers
Single tool 'add' has a basic schema with typed parameters (a, b as numbers) and descriptions, but falls short of production quality. No output schema documented. No error handling guidance. No parameter constraints (min/max for numbers). Tool does one thing (addition) but lacks depth expected for agent-ready tools. Naming is acceptable but generic.
Adds two numbers together
Tool description is 18 characters ('Adds two numbers together'), below 20-char minimum. LLM cannot determine when/why to select this tool vs alternatives.
No output schema documented. LLMs cannot plan downstream operations or extract results. Rubric baseline: 100% of A+ tools have documented return types.
Numeric parameters lack constraints (min/max). Unbounded 'a' and 'b' allow LLMs to pass absurd values (e.g., 1e308) that may cause overflow or unexpected behavior.
No error handling guidance. If addition fails (e.g., type coercion error), LLM receives no recovery path. Rubric requires: 'Error responses must tell the LLM what to do next.'
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 39 | 2026-07-28+ | v2 |
Tool is trivial (arithmetic). Real-world MCP servers expose domain-specific operations (API calls, data queries, state mutations). This demo lacks practical utility for agent evaluation.