NASA MCP provides 5 well-named, read-only tools with complete input schemas and reasonable descriptions. All tools follow verb_noun naming (get_*, search_*). Schemas are properly typed with required/optional fields clearly marked. However, descriptions lack depth, most are 100-150 chars and omit guidance on WHEN to use each tool vs alternatives, dependency hints, or error recovery. Output schemas are not documented, forcing LLMs to infer response structure. Parameter descriptions are adequate but generic (e.g., 'Date in YYYY-MM-DD format' lacks range context). No error handling guidance, no pagination support despite potential for large result sets (search_nasa_images). Security is sound (no secrets in params, read-only operations), but composition could be stronger, search_nasa_images and get_apod both return image data but with different structures.

• Add output schema documentation to each tool. E.g., 'Returns: {date: string, title: string, explanation: string, url: string, hd_url?: string, media_type: string, copyright?: string}' in the tool definition or a separate outputSchema field.
• Expand descriptions to 150-250 chars with WHEN-to-use guidance. E.g., 'Get the NASA Astronomy Picture of the Day with explanation. Use for astronomy/space education queries. Optionally specify a date (1995-06-16 to today); defaults to today.'
• Replace example values with enum constraints. E.g., camera parameter: use enum ['FHAZ', 'RHAZ', 'MAST', 'CHEMCAM', 'MAHLI', 'MARDI', 'NAVCAM'] instead of 'e.g. "FHAZ"'.
• Add error handling with actionable messages. Catch HTTP errors and return structured errors: {error: 'invalid_date', message: 'Date must be YYYY-MM-DD between 1995-06-16 and today', suggestion: 'Try a recent date like 2024-01-15'}.
• Add pagination support to search_nasa_images: accept offset/limit and return {results: [...], total: number, has_more: boolean} so LLM can fetch additional pages.