MCP server for querying nutritional information from the Nutritionix API. Provides tools to search for food items and retrieve detailed nutrient summaries.
The Nutritionix MCP server has moderate definition quality with critical gaps. Both tools have descriptions present but lack comprehensive input/output schema documentation. The `search_food` tool has a single string parameter with a description, but the `get_nutrient_summary` tool is poorly named (confusing verb+noun), and neither tool documents its output schema in the code, critical for LLM composition. The descriptions are adequate (100-150 chars) but lack WHEN to use each tool and what the LLM should expect back. No error handling guidance is present, no parameter constraints (enums, ranges), and no documentation of field dependencies between tools. Code review reveals a typo in search_food.py ('http_clien' vs 'http_client'), indicating lack of verification. Overall, the definitions are sufficient for basic use but would not pass production code review for an agentic system.
Get a summary of nutritional information for a food item. This MCP tool queries the Nutritionix API to fetch basic serving and measurement information for a given food query. It returns key details about the food item including serving size, weight, and available alternate measurement units.
Search for food items in the Nutritionix database. This MCP tool queries the Nutritionix API's instant search endpoint to find matching food items for a given search term. It returns a list of standardized food names from both common and branded food databases.
Output schemas not documented. get_nutrient_summary returns a dict with fields (food, serving_quantity, serving_unit, etc.) but this is not declared in the tool registration or docstring. LLMs cannot plan downstream tool calls without knowing what fields will be returned.
Tool naming is ambiguous. 'get_nutrient_summary' conflates two responsibilities: (1) parsing a natural-language food query and (2) fetching summary data. A clearer name would be 'get_food_nutrition' or 'lookup_food_nutrients'. The current name does not clearly convey that it accepts natural language input (e.g., '1 large banana') rather than a simple food identifier.
No error handling guidance. Both tools return vague fallback strings (e.g., 'Could not Fetch Food Item Information') without actionable recovery hints. LLMs cannot determine if the error is retryable, user-fixable, or fatal. Per the rubric, error responses must tell the agent what to do next.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | F | 44 | 2026-07-28+ | v2 |
| 2026-03-09 | F | 36 | - | v1 |
Parameter constraints missing. The 'query' parameter in both tools accepts free-form strings with no guidance on length, format, or examples of valid inputs vs invalid ones. For 'search_food', should it accept 'apple' or 'Red Delicious apple' or both? Undocumented constraints invite LLM hallucination.
No result pagination or limits documented. search_food returns a list but does not specify: (a) how many items are returned, (b) if pagination is supported, (c) what happens if 1000 matching foods exist. The docstring should state 'Returns up to 20 results' and offer offset/limit parameters if the API supports it.
Typo in source code: nutritionix_mcp/tools/search_food.py imports 'http_clien' instead of 'http_client'. This will cause a runtime import error and indicates the code has not been tested or reviewed.