MCP server that allows to search web using Brave API
Single tool with adequate schema structure but critical gaps in descriptions and parameter documentation. The tool name is action-oriented (web_search), which is good, but descriptions lack depth and parameter descriptions are entirely missing, violating foundational LLM optimization principles. The tool counts as 1 total; evaluation below shows individual scores average to 48.
Search web
Tool description is extremely minimal (9 characters: 'Search web'). This fails the 10-1024 character guideline and provides no context for LLM selection. Does not explain WHEN to use it, WHAT it returns, or any prerequisites.
ALL input parameters (count, country, search_lang, freshness) lack descriptions in the schema. Parameter names alone are ambiguous, 'count' could mean number of results, API limit, or timeout count. 'search_lang' is not obviously different from 'country'. LLMs cannot infer meaning without explicit descriptions.
Output schema is not documented. The tool returns JSON but there is no description of what fields to expect (e.g., results array structure, field names, types). LLMs cannot plan downstream tool calls or extract data without knowing response shape.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | F | 7 | <=2025-11-25 | v2 |
| 2026-03-09 | F | 38 | - | v1 |
Parameter 'count' is defined as string type with regex pattern '^\d+$', but this is poor UX and schema design. Numeric counts should be integer type, not string. The code comment acknowledges 'integer is not supported in MCP Inspector v0.3.0', this is a workaround for tooling limitations, not a valid design choice. Modern MCP supports integers.
No validation error messages guide the LLM. If freshness regex fails, the error returned is generic ('Invalid arguments: ...'). Should be actionable: 'Invalid freshness: must be YYYY-MM-DDtoYYYY-MM-DD format, e.g., 2024-01-15to2024-01-31'.
No documentation of result limits or pagination. The description does not state how many results are returned by default, whether pagination is supported, or how to fetch more results if needed. Returning unbounded result sets risks context window exhaustion.
freshness parameter uses regex pattern '^\d{4}-\d{2}-\d{2}to\d{4}-\d{2}-\d{2}$' but the schema does not document this format constraint in the description. LLMs cannot read JSON Schema pattern fields, they rely on description text.