Two tools with complete input schemas and descriptions. chat tool has comprehensive parameter documentation (messages, model, temperature, max_tokens, top_p, stop, system, include_reasoning) with proper types and enums. list_models is minimal but valid. Descriptions are adequate (chat: 156 chars, list_models: 60 chars) but lack actionable guidance on when to use each tool or error recovery paths. No output schemas documented. Parameters lack constraint details (e.g., temperature range 0-2 stated in description but not in schema; no min/max for max_tokens). Error handling not visible in tool definitions.

• Document output schema for chat(): return type should specify fields like 'content' (string), 'reasoning' (optional string if include_reasoning=true), 'model' (string), 'usage' (object with prompt_tokens, completion_tokens). This enables LLMs to extract and chain results.
• Document output schema for list_models(): specify return structure (e.g., {default_model: string, aliases: {[key: string]: string}}) so LLMs know how to parse the response.
• Add error handling guidance to chat description: 'Returns error if API key invalid, model not found, or rate limited. Retryable errors include timeout and temporary service unavailability. Non-retryable: invalid API key (check OPENAI_COMPATIBLE_MCP_API_KEY env var), unknown model (call list_models first).'
• Formalize parameter constraints in JSON Schema: add 'minimum': 0, 'maximum': 2 to temperature; add 'minimum': 1 to max_tokens; add 'minLength': 1 to messages array.
• Enhance list_models description: 'Lists the configured default model and friendly model aliases. Call this first to discover which models are available before invoking chat(). Returns {default_model, aliases}.' (currently 60 chars, should be 120-180).

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

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