This server has severe definition quality issues. Tool names lack action verbs (sampling, logging, toolInvocation, health are not verb_noun patterns). Descriptions are present but generic and lack context on WHEN to use each tool or what happens. Schemas exist but are incomplete: parameters lack proper type constraints, enums are missing for constrained inputs (e.g., log level should be enum), and output schemas are defined in Pydantic models but not exposed in tool registration. The 'sampling' tool is particularly problematic, it appears to be a deprecated MCP pattern (server-initiated sampling) rather than a proper agent tool. No error handling guidance, no recovery paths, no security considerations documented.

• Rename tools to verb_noun pattern: 'sampling' → 'invoke_model_sampling', 'logging' → 'send_log_message', 'toolInvocation' → 'execute_tool', 'health' → 'check_health'.
• Expand descriptions to 100-200 chars explaining WHAT, WHEN, and WHAT-RETURNS. Example: 'invoke_model_sampling: Generate text completions from a specified model. Use when you need to sample from a language model. Returns generated text and token count.'
• Add enum constraint to logging level parameter: {"type": "string", "enum": ["info", "warning", "error"], "description": "Log level (one of: info, warning, error)"}.
• Document output schemas in tool definitions (not just Pydantic models). Include field descriptions and types so LLMs know what to expect.
• Add error handling guidance: document which errors are retryable, which require user input, and what the LLM should do next (e.g., 'If model not found, call list_available_models first').

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

• Relies on Logging (deprecated) - log to stderr or use OpenTelemetry