Server provides 3 tools with reasonable input schemas and descriptions, but has significant gaps in parameter documentation, output schema clarity, and error handling guidance. Tool names follow verb_noun convention (generate_image, get_image_data, reload_config), but descriptions lack specificity about return structure and error recovery paths. Parameter descriptions exist but are inconsistent in detail, some provide format guidance ('Format: provider:style'), while others are sparse. No documented output schemas, no error recovery guidance for LLM clients. Server operates over STDIO+HTTP but schema definitions are embedded in code without formal registration visible. Overall falls into 'Fair' category with noticeable gaps.

• Document output schema for generate_image: {image_id: string, provider: string, url: string (if available), dimensions: {width, height}, created_at: ISO8601, expires_at: ISO8601?, base64_data?: string}. This enables downstream tools (like get_image_data) to reference images reliably.
• Add formal enum constraints to generate_image. Replace 'Leave empty to use default' with enum: ['hunyuan', 'openai', 'doubao', '']. Replace style/resolution free-text with provider-specific enums or documented format rules (e.g., resolution: '^\d+x\d+$').
• Expand get_image_data description to 100-150 chars explaining: 'Retrieve base64-encoded image data for a previously generated image. Use when you need to embed the image directly in a response or re-serve it without calling generate_image again.'
• Add error handling guidance: 'If generation fails, response includes error_code (invalid_prompt, api_rate_limit, invalid_provider, unsupported_resolution) and next_step (try_different_prompt, wait_and_retry, contact_provider). Retry only if next_step='wait_and_retry'.'
• Document idempotency: 'Calling generate_image twice with identical params will generate a new image each time (non-idempotent). Save the image_id from the first call if you need to retrieve the same image. Design does not support conditional/cached generation.'
• Add permission-gate documentation to reload_config: 'This tool reloads environment configuration and requires admin-level permission. Verify caller identity before executing. Log all configuration reloads for audit compliance.'

Inferred effective spec: 2025-06-18+.

• Relies on Stateful initialize / Mcp-Session-Id (removed; protocol is stateless) - make each request self-contained