A Cloudflare Workers-based MCP gateway that provides authentication, token exchange, and access to remote tools via OpenAPI specifications. Includes an agent web interface and remote tools worker for executing customer, order, and report operations.
Three tools with minimal descriptions (10-30 chars) and basic schemas. All tools are READ_ONLY, which is positive for safety, but descriptions lack context on WHEN to use each tool, WHAT they return, and HOW they differ. Schemas present but sparse, no enums, no min/max constraints, no pagination guidance. Parameter descriptions are generic ('Customer email address', 'Customer identifier', 'Date range'). No output schema documentation visible. No error handling guidance. Naming is acceptable (verb_noun pattern) but descriptions fall well below the 50-200 char baseline for LLM-optimized tools.
Find customer by email
Get latest order by customer id
Get sales summary report
Descriptions are critically short (10-30 chars). Rubric baseline is 50-200 chars for LLM-optimized tools. Current descriptions lack WHEN to use, WHAT is returned, and dependencies.
No output schema documented. LLMs cannot plan downstream calls or extract required fields without knowing response structure.
Parameter descriptions are generic and lack format/constraint guidance. E.g., 'Date range for the report' does not specify format (ISO 8601? natural language? 'last_7_days'?), valid values, or examples.
No error handling guidance. If customer.lookup fails (not found, invalid email), the response should guide the LLM to retry with search or ask the user.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 43 | 2026-07-28+ | v2 |
No pagination or result-limiting guidance. If report.salesSummary returns thousands of rows, context window is exhausted. Rubric baseline: cap results at 20-50 and document in description.