{"id":"qdrant-api","name":"Qdrant","homepage":"https://qdrant.tech","repo_url":"https://github.com/qdrant/qdrant","category":"database","subcategories":["vector-database","ai-infrastructure","semantic-search","rag"],"tags":["qdrant","vector-database","embeddings","semantic-search","rag","rust","open-source","mcp-server","hybrid-search","sparse-vectors","grpc","fastembed"],"what_it_does":"Qdrant is a high-performance open-source vector database and search engine written in Rust, designed specifically for AI and machine learning workloads. It stores high-dimensional embedding vectors alongside arbitrary JSON payloads and enables fast approximate nearest neighbor (ANN) search with rich metadata filtering. Qdrant supports dense vectors, sparse vectors (BM42), and multi-vector configurations — enabling hybrid search that combines semantic similarity with keyword relevance in a single query. Available as self-hosted (Docker/Kubernetes) or fully managed via Qdrant Cloud, with an official MCP server for direct AI agent memory integration.","use_cases":["RAG pipelines with semantic document retrieval and rich metadata filtering (date ranges, categories, user IDs)","Hybrid search combining dense embedding vectors with sparse BM42 keyword vectors in a single query","AI agent persistent semantic memory via the official MCP server (qdrant-store / qdrant-find tools)","Recommendation systems using vector similarity plus business logic payload filters","Real-time vector search at scale with on-disk HNSW indexing for datasets exceeding RAM","Multi-tenant vector databases using collection-level isolation or payload-based filtering with named vectors"],"not_for":["Teams that want relational data with complex multi-table joins (use PostgreSQL instead)","Pure keyword search without semantic understanding (use Elasticsearch or Typesense)","Workloads requiring ACID transactions across multiple documents","Teams without any DevOps capability who need a fully managed zero-touch database"],"best_when":"You need a high-performance, resource-efficient vector database with first-class MCP server support, the ability to self-host for free, or a competitively priced managed cloud option — especially for RAG, semantic search, or agent memory.","avoid_when":"Your workload is purely keyword-based full-text search, or you need relational database guarantees like foreign keys and complex joins.","alternatives":[{"id":"pinecone-api","reason":"Fully managed with no self-hosting option; simpler but pricier at scale"},{"id":"weaviate-api","reason":"Built-in vectorizer modules and GraphQL API; more opinionated pipeline"},{"id":"mcp-server-qdrant","reason":"The MCP-only interface for Qdrant without direct REST/gRPC access"}],"af_score":84.7,"security_score":75.0,"reliability_score":null,"package_type":"mcp_server","discovery_source":["github"],"priority":"low","status":"evaluated","version_evaluated":"current","last_evaluated":"2026-03-01T09:50:06.108003+00:00","performance":{"latency_p50_ms":5,"latency_p99_ms":30,"uptime_sla_percent":99.9,"rate_limits":"No hard rate limits on self-hosted; Qdrant Cloud scales limits with plan size","data_source":"llm_estimated","measured_on":null}}