Loading ToolBench

Reading the current corpus…

mcp-themealdb MCP server · C · ToolBench

mcp-themealdb

TheMealDB MCP. Open meal/recipe database — search recipes by name, look up full recipes with ingredients and instructions, filter by ingredient/category/cuisine, and list available categories, cuisines, and ingredients. Keyless (uses the public test key "1").

2026-07-28+4 tools

61/100

Definition Quality81

Protocol Readiness40

Supportability43

Spec posture

Inferred effective spec 2026-07-28+·current baseline 2026-07-28

Static source inference · medium confidence

Current-spec patterns detected

Summary

TheMealDB MCP has well-structured tool definitions with clear naming (verb_noun pattern) and complete input schemas. All 4 tools have descriptions (194 - 250 chars, within baseline 34 - 392 range) and typed parameters with descriptions. However, output schemas are not documented, LLMs cannot predict response structure for planning. Error handling is minimal (generic error objects). Parameter descriptions include example values (e.g. 'chicken_breast', 'Italian'), which LLMs may reuse literally. No pagination metadata (total_count, next_cursor) despite tools returning lists. Missing tool annotations (readOnlyHint, idempotentHint) despite all being read-only operations.

Tools (4)

filter_mealsread onlysource verified78/100

Filter recipes by ONE of: main ingredient, category, or cuisine (area). Returns a light list (id, name, thumbnail) — use get_meal for full details. Priority if multiple given: ingredient > category > area. Keyless.

get_mealread onlysource verified83/100

Look up a single full recipe by its TheMealDB meal id — complete (untruncated) instructions, full ingredient list, tags, source URL, and YouTube link. Keyless.

list_optionsread onlysource verified80/100

List the available filter values in TheMealDB: all categories, all cuisines (areas), or all ingredients. Use the returned values with filter_meals. Keyless.

search_mealsread onlysource verified83/100

Search recipes by name in TheMealDB. Returns matching meals with ingredients, cuisine, category, thumbnail, truncated instructions, and a YouTube link. Keyless.

Issues (5)

HIGHtool ↗

Output schemas not documented. LLMs cannot predict response structure (fields, types, nesting) for downstream planning or data extraction.

search_mealsget_mealfilter_mealslist_options

MEDIUMconstrained-input ↗

Example values in parameter descriptions (e.g. 'chicken_breast', 'Italian', '52771') risk LLM literal reuse. Use enums or format constraints instead.

search_mealsget_mealfilter_meals

MEDIUMpaginated-result ↗

No pagination metadata (total_count, next_cursor, has_more). search_meals and filter_meals cap results at 25 and 50 but don't signal truncation to LLM.

search_mealsfilter_meals

LOWtool-annotations ↗

No tool annotations. All 4 tools are read-only and idempotent but lack readOnlyHint/idempotentHint metadata for agent safety reasoning.

search_mealsget_mealfilter_meals

Recommendations

Document output schemas for all tools. Example for search_meals: {type: 'object', properties: {count: {type: 'integer'}, meals: {type: 'array', items: {type: 'object', properties: {id, name, category, area, thumbnail, ingredients: [{name, measure}], instructions, youtube}}}}}.
Replace example values with formal constraints. For filter_meals, use enum for 'type' (already done for list_options); for ingredient/category/area, add pattern or minLength/maxLength.
Add pagination support: include 'total_count' in search_meals and filter_meals responses; document result limits (25, 50) in tool descriptions.
Add tool annotations in callTool dispatch or via MCP tool metadata: readOnlyHint: true and idempotentHint: true for all 4 tools.
Enhance error messages with recovery hints. E.g. 'Meal not found. Try search_meals() with a partial name or list_options(type="categories") to explore available filters.'

Score history

Overall score trend

First recorded score · v2 rubric

61/100

Scored	Grade	Overall	Spec posture	Rubric
2026-09-23	C	61	2026-07-28+	v2