The Wordware MCP (Master Control Program) server allows you to run your Wordware apps locally, with interactive setup.
Critical gaps across all dimensions. Tool 1 (tools/list) has a minimal description (47 chars) and sparse schema with only a cursor parameter lacking description. Tool 2 (dynamic tools) is inferred from code comments rather than explicit registration, tool names, descriptions, and schemas are generated at runtime from an external API, making them invisible in static analysis. The codebase contains extensive schema patching logic (transformInputSchema, transformResponseFormat) suggesting the Wordware API returns malformed schemas that the MCP server must repair. No error handling guidance, no output schema documentation, no parameter validation rules, and no security controls visible. The server acts as a thin proxy to an external API without adding MCP-compliant structure.
Tools are dynamically registered from the Wordware API response. Each tool name is formatted to match MCP compatibility (^[a-zA-Z0-9_-]{1,64}$) and executed via the Wordware RPC endpoint.
Fetches available tools from the Wordware API
Dynamic tools are inferred, not explicitly registered. Tool names, descriptions, and schemas come from runtime API responses, not visible in source code. Per HARD SCORING RULES, inferred tools cap at 50 overall; here they score 0 because no schema is statically verifiable.
tools/list description is 47 characters ('Fetches available tools from the Wordware API'), below the 10 - 1024 baseline and lacks WHEN to use it, WHAT it returns, or any context for LLM selection. Per HARD SCORING RULES, descriptions under 20 chars score 0 - 20; this barely exceeds that threshold.
tools/list input schema has a 'cursor' parameter with no description. LLMs cannot infer whether cursor is a pagination token, a bookmark, or an opaque string.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 30 | 2025-06-18+ | v2 |
No output schema documented for either tool. LLMs cannot plan downstream calls or extract required fields (e.g., tool IDs, names, descriptions) without knowing the response structure.
Extensive schema patching logic (transformInputSchema, transformResponseFormat) indicates the Wordware API returns non-compliant schemas. The server repairs them at runtime, but this is fragile, undocumented, and suggests the tool definitions are not MCP-native. No error handling or recovery guidance for malformed API responses.