Qdrant MCP — Vector Search Engine for AI Agents

Available Tools

qdrant_store

Store documents with automatic embedding:

"Store this API documentation in the knowledge base under collection 'docs'"

qdrant_search

Semantic search across stored vectors:

"Find documents similar to 'authentication flow with JWT tokens'"

qdrant_delete

Remove specific entries:

"Delete all outdated documentation from the 'docs' collection"

qdrant_list_collections

View all collections and their stats:

"Show all vector collections and their sizes"

Use Cases

RAG Pipeline

Store project documentation, then retrieve relevant context for every query:

1. Index your docs: "Store all markdown files from ./docs into collection 'project-docs'"
2. Query with context: Agent auto-retrieves relevant docs before answering

Code Pattern Search

Store code patterns and search semantically:

"Find code similar to error handling patterns in the auth module"

Knowledge Management

Build a persistent knowledge base your agent can query:

"Remember that we chose PostgreSQL for JSONB support"
→ Stored as vector, retrievable by semantic search

Key Stats

• 22,000+ GitHub stars (Qdrant)
• Sub-millisecond search on millions of vectors
• Filtering + payload storage
• Real-time indexing
• Rust-powered performance

FAQ

Q: What is Qdrant MCP? A: Qdrant MCP is a Model Context Protocol server that connects AI agents to the Qdrant vector database for storing and searching embeddings, enabling RAG pipelines and semantic search.

Q: Is Qdrant free? A: Yes, Qdrant is open-source under Apache 2.0. Free self-hosted, with a managed cloud option available.

Q: Do I need to manage embeddings manually? A: The MCP server can auto-generate embeddings using configured models, or you can provide pre-computed vectors.