{"id":"galbaz1-video-research-mcp","name":"video-research-mcp","af_score":63.2,"security_score":54.8,"reliability_score":35.0,"what_it_does":"video-research-mcp is a Claude Code plugin that installs and configures MCP servers to enable video understanding/analysis, deep research with evidence grading, academic/paper search and extraction, web search with citations, and an optional knowledge store (Weaviate) for persistent semantic recall. It also describes standalone MCP server usage and related explainer/video/scene-generation MCP servers.","best_when":"You want an MCP-capable agent workflow that combines video analysis + citation-grounded research and optionally persistent semantic memory using Weaviate.","avoid_when":"You need formal API contracts (OpenAPI) or strong, documented reliability/error-handling semantics at the protocol level; also avoid when secrets/data cannot be sent to external AI providers or stored in third-party services.","last_evaluated":"2026-03-30T15:35:15.433000+00:00","has_mcp":true,"has_api":false,"auth_methods":["API keys via environment variables (GEMINI_API_KEY; optional WEAVIATE_API_KEY, COHERE_API_KEY, ELEVENLABS_API_KEY, OPENAI_API_KEY, YOUTUBE_API_KEY, S2_API_KEY)","MCP client authentication is effectively inherited from the MCP server process environment"],"has_free_tier":false,"known_gotchas":["Large local files (>20MB) are uploaded to Gemini File API for context-caching; agents may incur upload latency/cost and should avoid repeated uploads where possible.","Weaviate-backed semantic recall is optional; without configuration, recall falls back to keyword grep, which may reduce answer quality.","Some advanced capabilities depend on optional keys/services (YouTube Data API, Semantic Scholar, Cohere reranker, ElevenLabs/OpenAI/Sora, MLflow tracing).","The package relies on Claude Code plugin installation/assets plus MCP servers; integration may vary by client (Claude Code vs generic MCP client)."],"error_quality":0.0}