Qdrant

Vector storesApache-2.0

Qdrant stores vectors with attached payloads and returns the nearest matches to a query vector. It is written in Rust. You can run it locally, in a managed cloud cluster, as hybrid cloud on your own infrastructure, or embedded through Qdrant Edge. Official clients exist for Python, JavaScript and TypeScript, Rust, Go,.NET and Java, and the server exposes REST and gRPC APIs. An open-source MCP server gives an assistant a store and a find tool.

Pinout

  • MCP server: supported, verified
  • HTTP API: supported, verified
  • command line: not checked yet
  • Python SDK: supported, verified
  • TypeScript SDK: supported, verified
  • OpenAPI description: not checked yet
Where each pin comes from
InterfaceStatusSource we readRead on
MCP serverSupported, verifiedQdrant MCP (Open Source) page
HTTP APISupported, verifiedAPI and SDKs page, REST and gRPC sections
Command lineNot checked yet——
Python SDKSupported, verifiedAPI and SDKs page, Python row (pip install qdrant-client[fastembed])
TypeScript SDKSupported, verifiedAPI and SDKs page, JavaScript/TypeScript row (npm install @qdrant/js-client-rest)
OpenAPI descriptionNot checked yet——

Facts

Facts about Qdrant
LicenseApache-2.0. GitHub repository, About sidebar and README, read .
Latest GitHub releasev1.20.1. GitHub releases, latest, read .
PlatformsLocal quickstart, Qdrant Cloud, Hybrid Cloud, embedded Qdrant Edge. Documentation home, deployment and quickstart links, read .
Official clientsPython, JavaScript/TypeScript, Rust, Go, .NET, Java. API and SDKs page, read .
APIsREST (OpenAPI reference) and gRPC (protobuf definitions published). API and SDKs page, read .
MCP server toolsqdrant-store and qdrant-find. Qdrant MCP (Open Source) page, read .
Pricing, free cloud tierFree forever: 0.5 vCPU, 1GB RAM, 4 GB Disk. Pricing page, Free tier, read .
Paid cloud billingStandard tier is usage-based and billed hourly; the page lists no amounts. Pricing page, Standard tier, read .
Last commit. GitHub API, latest commit on the default branch of qdrant/qdrant, read .
FrameworksAgno (Framework Integrations table, read 2026-11-16); AutoGen (Framework Integrations table, read 2026-11-16); Camel (Framework Integrations table, read 2026-11-16); Cognee (Framework Integrations table, read 2026-11-16); CrewAI (Framework Integrations table, read 2026-11-16); DeepEval (Framework Integrations table, read 2026-11-16); DSPy (Framework Integrations table, read 2026-11-16); Haystack (Framework Integrations table, read 2026-11-16).

Call example

from qdrant_client import QdrantClient

client = QdrantClient(url="http://localhost:6333")

Limits

  • The MCP server page says the server has no built-in authentication, so a remote deployment needs a reverse proxy or a private network.

Strengths

  • The same project publishes official clients in six languages.
  • The open-source MCP server runs over stdio, SSE or streamable HTTP and needs no embedding provider key (FastEmbed).

Pricing

Pricing, free cloud tierFree forever: 0.5 vCPU, 1GB RAM, 4 GB DiskPricing page, Free tier, read 16 Nov 2026
Paid cloud billingStandard tier is usage-based and billed hourly; the page lists no amountsRead 16 Nov 2026

As written on Qdrant's own page on the read date. Prices change: check there before you pay.

  1. Embedded vector database library with Python, TypeScript and Rust SDKs.

    Vector stores
  2. Open-source vector database with Lite, Standalone and Kubernetes deployments.

    Vector stores
  3. Open-source vector database with hybrid search, SDKs and a built-in MCP server.

    Vector stores
  4. Open-source embedding database with Python and TypeScript clients and a CLI.

    Vector stores

Last check: . Method: public repository and docs read automatically, page and day written down. Next check due within 90 days. How we check

Is this your tool? Write to hello@agent-tools.directory from an address on its domain to correct a pin or a fact.