Single tool 'generate_image' has a complete input schema with proper Zod validation (prompt: string, aspect_ratio: enum). However, the tool description is in Turkish and lacks English clarity for international LLM use. The description does not explain WHEN to use this tool, what it returns, or error recovery paths. Parameter descriptions are minimal (e.g., 'Görsel için metin açıklaması' = 'Text description for image'). No output schema is documented, the tool returns an image URL but this is not formally specified. Error handling exists (retry logic, timeout) but error messages are not actionable for LLM recovery. The tool does one thing (generate image) which is good, but lacks the depth of documentation expected for production use.

• Translate all descriptions and parameter docs to English. Provide clear, LLM-optimized descriptions (50 - 200 chars) explaining WHAT the tool does, WHEN to use it, and WHAT it returns. Example: 'Generate an AI image from a text prompt using Freepik Flux-Dev. Specify aspect ratio (square, widescreen, etc.). Returns a URL to the generated image. Consumes API credits.'
• Document the output schema explicitly. Specify that the tool returns an object with fields like {url: string, task_id: string, status: string} or similar. Include field types and meanings.
• Enhance error messages to be actionable. Instead of 'Maksimum yeniden deneme sayısına ulaşıldı', return 'API request failed after 3 retries. Check your FREEPIK_API_KEY environment variable and try again in 60 seconds.' This guides the LLM on recovery.
• Add a cost/quota warning to the tool description. Inform users and agents that each call consumes credits. Consider adding a dry_run parameter (boolean, optional) to preview the request without consuming quota.
• Improve parameter descriptions with format hints. For aspect_ratio, explain: 'Image aspect ratio. Use square_1_1 for social media, widescreen_16_9 for presentations, portrait_2_3 for mobile. Defaults to square_1_1 if omitted.'