Foundation Foods MCP server demonstrates solid tool definition quality with well-structured schemas, clear naming conventions, and comprehensive parameter validation. All three tools follow a verb_noun pattern (search_*) with appropriate input constraints. However, descriptions are moderately detailed but lack depth on error handling, use cases, and recovery guidance. Output schemas are properly typed but response documentation in code comments is minimal. Tool composition is reasonable but shows some redundancy, three tools perform nearly identical searches with only nutrient-inclusion varying, which could confuse LLM tool selection.

• Consolidate the three search tools into two: (1) 'search_foundation_foods' (basic, returns food metadata), and (2) 'search_foundation_foods_nutrients' (with optional nutrients_to_include array). Remove the simplified variant, it duplicates the customizable variant without clear added value.
• Document the DefaultNutrients enum explicitly in all tool descriptions. E.g., 'Default nutrients returned: protein, fat, carbohydrates, fiber, vitamins_minerals. Call with nutrients_to_include: ["calcium", "iron"] to customize.'
• Add error handling guidance to all tool descriptions. E.g., 'Returns empty array if no foods match. Partial matches supported (e.g., 'cheddar' matches 'Cheese, cheddar, sharp').'
• Include the output response structure in tool descriptions or as a separate schema document. Specify: 'Returns an object {foods: [{id, name, brand?, nutrients?: [...]}], total_count, next_cursor?}' so LLM knows what to expect.
• Clarify which nutrients are searchable via 'nutrients_to_include'. Provide a discoverable reference (e.g., 'Valid nutrient names: protein, fat, carbohydrates, fiber, calcium, iron, vitamin_a, vitamin_c, zinc'). Consider adding a separate 'list_available_nutrients' tool if the list is large or dynamic.
• Add 'When to use' section to descriptions: 'Use search_foundation_foods for product discovery. Use search_foundation_foods_nutrients when you need specific nutrient values (e.g., calculating macros for a recipe).'

Inferred effective spec: 2026-07-28+.

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