Loading ToolBench
An MCP server for context management, validation, and Cursor IDE integration
This MCP server has critically deficient tool definitions. Only 2 tools are exposed (validateResponse, healthCheck), both with minimal schemas and vague descriptions. validateResponse accepts a generic 'body' parameter with no type constraint or validation guidance. healthCheck takes no parameters. Neither tool has documented return schemas or error handling. The server implements basic HTTP transport but the tool interface fails to meet production standards for LLM interaction: descriptions are too short to guide selection, parameters lack specificity, and output structure is undocumented. The codebase shows competent Express/TypeScript infrastructure, but the MCP surface is underspecified.
Checks the health status of the Cursor IDE integration
Validates response from Cursor IDE integration
validateResponse parameter schema is severely underspecified. 'body' is typed as 'object' with no nested schema, no property descriptions, and no validation constraints. LLMs cannot infer what fields are expected or how to construct the request.
Tool descriptions are too brief (18-50 characters) to enable effective LLM selection. 'Validates response from Cursor IDE integration' lacks WHAT it validates, WHEN to call it, and WHAT the output structure is. Cannot reach 200+ character baseline for production tools.
No documented output schemas for either tool. LLMs cannot plan downstream operations or extract relevant response fields. Returns are completely opaque to the agent.
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | F | 28 | <=2025-11-25 | v2 |
| 2026-03-09 | F | 25 | - | v1 |
healthCheck has no parameters but the tool name is generic and does not follow verb_noun convention clearly ('health_status_check' or 'get_health_status' would be clearer). No description of what system is being checked or what health metrics are returned.
No error handling guidance. If validation fails or health check encounters an issue, no error schema or recovery suggestions are documented. LLMs have no actionable next steps on failure.
validateResponse mixes concerns: 'validates response from Cursor IDE integration' suggests it may perform multiple checks or side effects, but the single-parameter design and lack of output schema make its purpose unclear. No indication of what 'validation' means (schema validation, linting, syntax check, semantic verification).