A Cloudflare Workers-based MCP server with OAuth 2.0 authentication and authorization flow demonstration
This server has a single tool 'add' with critical quality gaps. The tool is explicitly registered in src/index.ts with Zod schema validation (z.number() for both 'a' and 'b'), but the implementation lacks essential documentation and descriptions. The tool name 'add' is generic and does not follow verb_noun convention (should be something like 'sum_numbers' or 'calculate_sum'). Most critically, parameter descriptions are null (as shown in the Input schema), and there is no tool description provided. The output is hardcoded to return text content without documenting the expected response structure for LLMs. Error handling is absent, no guidance on what happens with invalid inputs, edge cases (overflow, NaN), or recovery paths. The server is built on a custom framework (Hono + agents/mcp) that uses the @modelcontextprotocol/sdk, but the tool definition quality is far below production standards.
Missing tool description, tool 'add' has no description at all, making it impossible for LLMs to determine when or why to select it or what it returns.
Missing parameter descriptions, both parameters 'a' and 'b' are shown with description: null in the Input schema. LLMs cannot infer parameter meaning or constraints from names alone.
Generic tool name 'add' does not follow verb_noun convention and is ambiguous in context. Does it add to a list? Add numbers? Add users? Consider 'sum_numbers' or 'calculate_sum'.
No documented output schema, the tool returns hardcoded text content with no schema documentation for what fields LLMs should expect or how to chain to downstream tools.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | F | 26 | 2026-07-28+ | v2 |
| 2026-03-09 | F | 6 | - | v1 |
No error handling or input validation guidance, LLMs have no recovery path if invalid inputs are passed (e.g., non-numeric strings, NaN, Infinity, overflow). No guidance on what to do on failure.
No parameter constraints or format guidance, no minimum/maximum bounds, no description of expected numeric range, no guidance on precision or edge cases.