Mcp Qdrant
Give your coding agent read-only visibility into Qdrant collections so it can debug RAG retrieval, counts, and vector search without mutating vectors.
Overview
mcp-qdrant is a Build-phase MCP server that provides read-only Qdrant collection inspection, point browse, count, and vector search for AI agents.
What is this MCP server?
- List collections and fetch collection info/metadata
- Browse points and get document counts read-only
- Vector search against existing collections
- Read-only guard—no destructive vector index operations via MCP
- npm stdio package @infoinlet/mcp-qdrant v0.1.1
- Server version 0.1.1 published as @infoinlet/mcp-qdrant on npm
- Documented tools span collections, info, browse points, count, and search
- stdio MCP transport per 2025-12-11 server schema
What problem does it solve?
Your RAG stack misbehaves in Qdrant but you keep copying payloads and query JSON into chat because the agent cannot see the vector store directly.
Who is it for?
Indie builders iterating on embeddings and retrieval who want agent-side debugging of Qdrant with a read-only MCP boundary.
Skip if: Operators who need to create collections, upsert vectors, or rebuild indexes entirely through the agent without admin tools.
What do I get? / Deliverables
After install, the agent can list collections, inspect points, count records, and run search probes against Qdrant without write access.
- Read-only collection and point inspection from the agent
- Search and count operations for retrieval validation
- Faster RAG troubleshooting without manual UI context switching
Recommended MCP Servers
Journey fit
Vector stores are wired while you build agent features and RAG pipelines, before launch analytics or production incident response. Qdrant access is agent-tooling: it supports embeddings workflows and retrieval debugging, not generic CRUD app code.
How it compares
Read-only Qdrant MCP bridge for RAG debugging, not an embedding pipeline or vector ETL skill.
Common Questions / FAQ
Who is mcp-qdrant for?
Solo developers building agent or SaaS features on Qdrant who need Claude Code or Cursor to inspect retrieval data safely.
When should I use mcp-qdrant?
Use it while building RAG or semantic search when you need to verify collection contents, counts, or search results during integration and tuning.
How do I add mcp-qdrant to my agent?
Add @infoinlet/mcp-qdrant as a stdio MCP server in your agent config and provide Qdrant host URL plus API credentials with read-appropriate permissions.