Search and RAG MCP servers connect AI agents to semantic search engines, vector databases, and retrieval-augmented generation pipelines. These tools are the infrastructure layer that lets language models answer questions grounded in your own data — not just their training data.

The category covers two types: web search tools (Brave Search, Tavily, SerpAPI) that give agents access to current information, and vector store tools (Pinecone, Weaviate, Qdrant, Chroma, Milvus) that enable semantic retrieval over private document collections. Both are essential for production AI applications.

For RAG workloads, prioritize tools with chunking strategies, metadata filtering, and hybrid search (keyword + semantic). For web search, look for tools that return structured results with citations — raw HTML dumps are much harder for agents to work with than clean snippets.