Loading ToolBench

Reading the current corpus…

movie-mcp MCP server · D · ToolBench

movie-mcp

An MCP server that provides tools to search and query a MongoDB movie database (sample_mflix) with filtering, sorting, and rating options.

2026-07-28+STDIO1 tools

58/100

Definition Quality77

Protocol Readiness50

Supportability33

Spec posture

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

Static source inference · medium confidence

Current-spec patterns detected

Summary

The server defines a single tool 'find_movies' with a comprehensive schema and reasonable descriptions. However, there are notable gaps in parameter descriptions, missing error handling guidance, and the tool lacks any hint annotations. The schema is well-structured but descriptions could be more LLM-optimized. Parameter descriptions vary in quality, some are detailed (e.g., 'sort_by' with field examples), others are generic (e.g., 'genres' just says 'List of genres'). Output schema is not documented. Error handling provides no recovery guidance.

Tools (1)

find_moviesread onlysource verified77/100

Finds movies based on a variety of criteria, with options for sorting and limiting results.

Issues (6)

HIGHschema-documentation ↗

Output schema not documented. Tool description states 'Finds movies...' but does not specify what fields are returned, their types, or when pagination applies.

find_movies

MEDIUMconstrained-input ↗

Parameter descriptions lack clarity on constraints and formats. E.g., 'sort_by' says 'Field to sort results by (e.g., "imdb.rating", "year", "title", or short keys: "imdb", "metacritic")', example values should be replaced with enum constraints or a reference to available fields. LLMs tend to reuse example values literally.

find_movies

HIGHrecovery-guide ↗

No error handling or recovery guidance in tool definition. Errors like 'MongoDB connection failed' or 'empty result set' have no actionable next steps for the LLM. Pattern recommends: 'Try narrowing filters' or 'Call search with partial title.'

find_movies

MEDIUMtool-annotations ↗

Missing tool annotations. Tool is READ_ONLY (safe to retry, no side effects) but this hint is not declared in the schema via readOnlyHint or equivalent. LLMs cannot infer retry safety without explicit hints.

Recommendations

Document the output schema explicitly in the tool description. Include fields like 'title', 'year', 'plot', 'imdb.rating', 'genres' with their types and an example response snippet.
Replace example values in 'sort_by' parameter with an enum or a formal list of valid values. E.g., sort_by must be one of: 'title', 'year', 'imdb.rating', 'metacritic', 'tomatoes.viewer.rating', 'tomatoes.critic.rating'.
Add explicit error handling to the implementation: return 'No results found. Try broadening filters (remove genre restrictions or lower rating minimums).' when result set is empty. Return 'Invalid sort field. Must be one of: title, year, imdb.rating, metacritic, ...' when sort_by is invalid.
Add readOnlyHint or equivalent annotation to indicate the tool is read-only and always safe to retry.
Standardize list parameters to accept only array types, not union(list|string). Move the JSON parsing logic to the fastmcp framework level so the tool signature is clean.
Add a 'page_size' and 'page' or 'cursor' parameter to enable pagination. Cap limit at a reasonable maximum (e.g., 100) and document this in the description.
Expand parameter descriptions for 'genres', 'actors', 'directors', 'writers': include format guidance ('comma-separated names', 'case-insensitive'), and explain the AND logic ('all specified genres must match').
Add a note to the tool description: 'Returns a maximum of {limit} results (default 10, max 100 per call). If more results are needed, use pagination.' Include guidance on expected latency for large datasets.

Score history

Overall score trend

First recorded score · v2 rubric

58/100

Scored	Grade	Overall	Spec posture	Rubric
2026-09-22	D	58	2026-07-28+	v2