Brave Search MCP Server with HTTP Streamable transport - provides comprehensive web search, image search, video search, news search, autosuggest, and spellcheck capabilities using the Brave Search API
Argus demonstrates solid definition quality with consistently good tool naming (all verb-prefixed), comprehensive parameter descriptions, and well-documented input schemas. All 7 tools follow the verb_noun pattern (search_*, suggest, spellcheck). Descriptions are detailed and exceed the 194-char baseline average. Parameter schemas are complete with type definitions and descriptions for all parameters. However, output schemas are entirely undocumented, the code shows no explicit return type documentation in tool definitions. Additionally, error handling lacks recovery guidance (no 'try X instead' messages), and there is no mention of rate limiting, timeouts, or partial failure handling. Security posture is good (API keys injected via ContextVar, not exposed as parameters), but lacks explicit permission documentation and audit trail guidance.
Image search using Brave Image Search API. Performs image searches with filtering options optimized for visual content discovery.
News search using Brave News Search API. Searches for recent news articles with filtering by country, language, and freshness.
Video search using Brave Video Search API. Searches for video content with language and regional preferences.
Web search using Brave Search API (Data for Search). Performs comprehensive web searches with rich result types and advanced filtering options.
Web search with extra_snippets for AI/LLM using Data for AI key. Provides up to 5 additional excerpts per result, ideal for AI/LLM context. Falls back to regular search without extra_snippets if AI key fails.
Output schemas are completely undocumented. No return type definitions visible in code or docstrings for any of the 7 tools. LLMs cannot plan downstream operations or extract specific fields without knowing the response structure.
Error handling lacks recovery guidance. No error messages include 'try X instead' or actionable next steps. A failed API call returns raw errors without categorization (retryable vs user-fixable vs fatal). Violates recovery-guide pattern.
No documented result limits or pagination guidance in tool descriptions. While 'count' parameters exist with max values, descriptions do not explain pagination strategy or what happens when results exceed limits. Baseline best practice requires pagination documentation.
Inferred effective spec: 2026-07-28+.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | C | 69 | 2026-07-28+ | v2 |
| 2026-03-09 | F | 0 | - | v1 |
Spellcheck using Brave Spellcheck API. Checks spelling and provides corrections for search queries using the Data for Spellcheck plan.
Autosuggest using Brave Autosuggest API. Provides search query suggestions based on partial input using the Data for Autosuggest plan.
No timeout or rate-limit handling documented. External API calls (Brave Search) lack explicit timeout declarations or rate-limit guidance for agents. An agent in a retry loop could overwhelm the service without backoff strategy.
Parameter descriptions lack formal constraints. E.g., 'country' is described as '2-character country code' but no enum or regex pattern is declared. 'freshness' allows freeform strings matching 'pd|pw|pm|py|<date-range>' with no enum. LLMs will hallucinate invalid values.
No documented scope or permission requirements. Tools do not declare what Brave API plan/permissions they require (e.g., 'Data for Search', 'Data for AI'). Agents have no visibility into which API tier is needed or whether a tool call will fail due to subscription limits.