Ariadne MCP exposes 2 code-analysis tools with reasonable naming and moderate description quality. Both tools have clearly specified input schemas with type definitions and descriptions. However, there are notable gaps in output schema documentation, parameter constraints (enums, ranges), and error handling guidance. The tools follow a discovery + detail pattern (list_entrypoints → show_call_graph_neighborhood), which is compositionally sound, but lack pagination parameters and structured result limits. No evidence of permission gates, audit trails, or secrets management despite operating on code repositories. Tool descriptions are adequate (100-150 chars) but below the 'LLM-optimized' 50-200 char baseline. Parameter descriptions are present but sparse, lacking examples of valid format or constraint details.

• Document output schemas for both tools. For list_entrypoints: specify return structure (e.g. [{name, signature, call_tree_size, ref, file, line, is_test}]). For show_call_graph_neighborhood: specify caller/callee node structure with depth level, function name, file, line.
• Add pagination to list_entrypoints: include 'limit' (default 20, max 100) and 'offset' (default 0) parameters. Return total_count in response so LLM knows if results are truncated.
• Constrain numeric depth parameters: 'caller_depth and callee_depth must be 0 - 10. Depth 0 shows only the symbol itself; depth 1 shows direct callers/callees.'
• Clarify file/folder parameter format: 'Paths can be absolute or relative to repo root. Glob patterns supported (e.g. src/**/*.ts). If both files and folders specified, results filtered by either.'
• Add error recovery guidance to show_call_graph_neighborhood: 'If symbol_ref is invalid or symbol not found, try list_entrypoints first to verify the Ref. Common mistake: using source file path instead of Ref format (file_path:line#name).'
• Document that both tools are read-only (pattern:permission-gate). Add a note: 'These tools analyze code without modification. No side effects, safe to call repeatedly.'
• Add example symbol_ref formats to show_call_graph_neighborhood description to help LLM construct valid references: 'E.g., src/index.ts:42#getUserById or packages/core/src/api.ts:15#fetchData'.

Inferred effective spec: <=2025-11-25.

• Relies on Sampling (deprecated) - integrate directly with the LLM provider API