An MCP server that provides web search and local search capabilities using the Brave Search API, with integration examples for Google Gemini.
Two tools with good naming conventions and comprehensive descriptions, but schema documentation is incomplete and output structures are not formally documented. Tool names follow verb_noun pattern (brave_web_search, brave_local_search) appropriately. Descriptions are detailed and explain use cases well (ranging 150-250 chars, within the 10-1024 recommended range). However, input schemas lack formal type constraints and output schemas are entirely missing. No enum constraints despite opportunities (e.g., count parameter could have range constraints). Error handling exists but lacks recovery guidance. Tool definitions are explicitly visible in src/tools/definitions.ts (inferred from export), meeting the visibility requirement.
Searches for local businesses and places using Brave's Local Search API. Best for queries related to physical locations, businesses, restaurants, services, etc. Returns detailed information including: - Business names and addresses - Ratings and review counts - Phone numbers and opening hours Use this when the query implies 'near me' or mentions specific locations. Automatically falls back to web search if no local results are found.
Performs a web search using the Brave Search API, ideal for general queries, news, articles, and online content. Use this for broad information gathering, recent events, or when you need diverse web sources. Supports pagination, content filtering, and freshness controls. Maximum 20 results per request, with offset for pagination.
Output schemas not documented. The response structure (result fields, types, pagination info) is completely undocumented. LLMs cannot infer what fields to extract from results or plan downstream tool calls. Both tools return search results but the schema is invisible.
Input schema constraints missing. The 'count' parameter accepts 1-20 per description but lacks minItems/maxItems in schema. The 'query' parameter states '50 word max, 400 char max' but has no maxLength constraint. These must be formalized in JSON Schema, not just prose.
No error handling guidance. Handler in src/server.ts returns generic 'Unknown tool' and catch-all error messages with no recovery suggestions. When Brave API fails, LLMs get no actionable next steps (e.g., 'Try a different query', 'Rate limited, retry in 60s').
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | D | 52 | 2026-07-28+ | v2 |
| 2026-03-09 | F | 44 | - | v1 |
brave_local_search description mentions 'automatically falls back to web search if no local results found' but no tool parameter controls this behavior and no output field signals which mode was used. This creates ambiguity about what the LLM is actually calling.
No pagination documentation in output. brave_web_search has an 'offset' parameter supporting pagination (max 9), but response structure (whether it includes next_offset, total_count, has_more) is undocumented. Agents cannot reliably page through large result sets.