A comprehensive toolkit for data analysis, validation, and diagnostics with MCP support for dual transport (HTTP and Stdio)
This MCP server has critical definition quality issues. The 6 tools listed are mostly MCP protocol handlers (list_tools, handle_call_tool, list_resources, etc.) rather than domain-specific tools that expose analyst toolkit functionality. Tools 1-5 are MCP framework boilerplate with minimal descriptions. Tool 6 (rpc_handler) is a generic JSON-RPC dispatcher. None of the actual analyst toolkit capabilities (data analysis, validation, diagnostics) are exposed as properly defined MCP tools. Descriptions are generic protocol descriptions, not LLM-optimized explanations of what each tool does for users. Schemas are minimal (many tools have empty input {} or generic object params with no field documentation). No tool has proper parameter descriptions, type validation, or error recovery guidance. No indication of what the toolkit actually does or how users should interact with it.
Centralized tool dispatcher for the official MCP protocol (stdio) that routes tool calls to registered handlers
Standard MCP resources/templates/list handler that returns resource URI templates for config and golden templates
Standard MCP resources/list handler that returns available resources with timeout handling
Standard MCP tools/list handler that returns all available tools with their descriptions and input schemas
Standard MCP resources/read handler that reads and returns resource content by URI with timeout handling
HTTP JSON-RPC 2.0 dispatcher for FridAI that handles method routing, authentication, and metrics recording
No domain-specific tools exposed. The server lists only MCP protocol handlers (list_tools, handle_call_tool, list_resources, read_resource, rpc_handler) rather than exposing actual analyst toolkit capabilities like 'analyze_data', 'validate_dataset', 'run_diagnostics', etc. Users cannot call toolkit functions through MCP.
Descriptions are generic MCP protocol documentation, not LLM-optimized tool descriptions. E.g., 'Standard MCP tools/list handler that returns all available tools with their descriptions and input schemas' tells the LLM this is a protocol handler, not what it should use it for or when. Descriptions do not explain use cases, prerequisites, or what the user gains from calling the tool.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | F | 41 | 2026-07-28+ | v2 |
| 2026-03-09 | F | 0 | 2024-05-01+ | v1 |
Empty or overly generic input schemas. Most tools have empty input {} (list_tools, list_resources, list_resource_templates) or generic object params (handle_call_tool: arguments={type:object}, rpc_handler: params={type:object}). Generic 'object' types provide no field-level schema, no type enforcement, no validation guidance for LLMs.
No parameter descriptions. Parameters like 'name' (handle_call_tool), 'arguments', 'uri', and 'method' have minimal or no descriptions explaining what they control, what format they expect, or constraints. LLMs cannot infer parameter meaning from bare names, 'arguments' is ambiguous without context.
No output schemas documented. The source does not show what these tools return (fields, types, examples). LLMs cannot plan downstream calls or extract data without knowing the response structure. The rubric requires 100% of A+ tools to have documented return types.
No error handling guidance. Descriptions do not explain what can go wrong, when to retry, or what the LLM should do if a call fails. No mention of retryable vs fatal errors, no actionable recovery steps. E.g., if read_resource fails to find a URI, what should the LLM do next?
Generic naming. 'handle_call_tool' and 'rpc_handler' are generic action names that do not clearly convey what they do. 'handle_call_tool' suggests dispatch logic, not a user-facing capability. Better names would reveal the actual function (e.g., 'execute_analysis', 'validate_data', 'list_available_analyses').
No tool annotations. The server does not declare readOnlyHint, destructiveHint, or idempotentHint on any tool. The rubric rewards tool annotations as a current pattern (2026-07-28 spec). Even if tools are metadata-heavy, annotations help LLMs reason about safety and retry logic.