Starter MCP workspace for Claude Desktop, Cursor, Continue, and OpenAI-compatible gateway configuration.
Four tools with identical schemas and minimal differentiation. All have descriptions (10-20 chars baseline), but lack depth on WHEN to use each tool vs. alternatives. Input schemas present but generic, all five params optional with no enums, ranges, or validation hints. No output schema documented. Tools appear to generate configuration snippets, but the distinction between 'gateway_profile', 'client_config_snippets', 'smoke_test_plan', and 'routing_checklist' is unclear from descriptions alone. No error handling guidance. Naming follows verb_noun pattern weakly ('gateway_profile' is noun-heavy; 'client_config_snippets' is compound). Per-tool scores average 42.
Generate Claude Desktop, Cursor, and Continue config snippets for an MCP gateway starter.
Normalize an OpenAI-compatible gateway profile for MCP client setup.
Generate a short checklist before routing agent workflows through a compatible gateway.
Generate provider-neutral OpenAI-compatible smoke-test commands.
No output schema documented. LLMs cannot plan downstream actions or extract structured data from results. All four tools return JSON via toToolResult(), but the shape of that JSON is undocumented.
Descriptions are too brief (10 - 20 chars) and lack WHEN/WHY guidance. 'Normalize an OpenAI-compatible gateway profile' does not explain when to call this vs. client_config_snippets or smoke_test_plan. LLMs cannot disambiguate.
All five input parameters are optional with no defaults, enums, or validation hints. 'baseUrl' accepts any string marked .url() in Zod, but description says 'e.g. https://www.tken.shop/v1', example values in descriptions invite literal reuse by LLMs.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 41 | 2026-07-28+ | v2 |
No error handling or recovery guidance. If baseUrl is invalid or serverPath does not exist, what does the tool return? How should the LLM retry or correct the input?
Tool names lack clear action verbs. 'gateway_profile' and 'routing_checklist' are noun-heavy; 'client_config_snippets' is compound. Stronger names: 'generate_gateway_profile', 'generate_client_config', 'generate_smoke_tests', 'generate_routing_checklist'.