Getting Started with Qdrant Managed Cloud
This page takes you from no account to a connected, production-ready cluster. Each section gives the values you need and links to the page that covers the topic in full.
Before You Start
| You need | Details |
|---|---|
| A Cloud account | Sign up with email, Google, or GitHub at cloud.qdrant.io. Each user can own up to 5 accounts. See Account Setup. |
| Billing details | Free clusters need no credit card. Standard clusters need a credit card or an AWS, GCP, or Azure Marketplace subscription, set up on the Billing Details screen. If you have a custom payment agreement, create your account first, then contact support. See Billing & Payments. |
| A role that can create clusters | The Owner and Admin roles can. For custom roles, see Cloud RBAC. |
Choose a Cluster Type
| Free | Standard | Premium | |
|---|---|---|---|
| Resources | 1 node with 0.5 vCPU, 1 GB RAM, and 4 GB disk, shared | Dedicated, sized by you | Dedicated, sized by you |
| Nodes | 1 | 1 or more | 1 or more |
| Providers and regions | Limited choice | AWS, GCP, and Azure, broad choice | AWS, GCP, and Azure, broad choice |
| Upgrades | With downtime | Zero downtime on multi-node clusters with replication | Zero downtime on multi-node clusters with replication |
| Backups | Manual snapshots through the API | Scheduled backups and restores | Scheduled backups and restores |
| Uptime SLA | Not included | 99.5% | 99.9%, or 99.95% for multi-AZ |
| Also includes | Basic monitoring, logs, and alerting | Scaling, audit logging, and GPUs for indexing (AWS only) | SSO, private networking, customer-managed keys, and 24/7 support |
Free clusters are suspended after 1 week without use and deleted after 4 weeks of inactivity. You can upgrade a free cluster to a standard cluster at any time. For the full comparison, see Create a Cluster and Premium Tier.
Size Your Cluster
Estimate the RAM and disk for each collection, then sum them. For dense vectors, the base requirement is:
dense_size = points * replication_factor * dimensions * bytes_per_dim
bytes_per_dim is 4 for float32 (the default), 2 for float16, 1 for uint8, and 0.5 for turbo4. For example, one million 768-dimension float32 points with a replication factor of 2 need about 5.7 GB. Add the HNSW index, payloads, payload indexes, and quantized copies on top of that.
- Start with the estimate, not the maximum. You can scale RAM and CPU up or down later. Changing RAM or CPU restarts every node, which causes a short downtime if your collections are not replicated. Increasing only the disk size does not restart the nodes.
- Disk only grows. You can add disk space but not remove it, so add it in steps.
- Trade RAM for disk when cost matters. Quantization and the
coldmemory tier reduce memory use. Plan more disk per GB of RAM when you use them.
The full formulas for sparse vectors, payloads, and indexes are in Capacity Planning. To get a cost estimate, use the Pricing Calculator or the Sizing Calculator.
Create a Cluster
- In the Cloud Console, open Clusters and select + Create.
- Select Free or Standard.
- Select a cloud provider and region. Pick the region closest to your application.
- Set RAM, vCPU, and the number of nodes. Choose at least 3 nodes for production.
- Add disk space, and on AWS choose a disk speed tier.
- Review the configuration and price, then select Create.
Provisioning takes some time. After that, the cluster serves REST on ports 443 and 6333, and gRPC on port 6334. All options, including GPUs, Multi-AZ, and Hybrid Cloud environments, are described in Create a Cluster.
Create an API Key and Connect
Every request to a cluster needs a Database API key.
- Open the cluster and go to API Keys.
- Select Create, and enter a name.
- Set an expiration. The default is 90 days. An empty value means the key never expires.
- Keep the default manage/write permission, or choose read-only. To restrict the key to specific collections, use the Collections tab.
- Select Create and copy the key. It is shown only once.
Your cluster endpoint has the form https://<cluster-id>.<region>.<provider>.cloud.qdrant.io. Requests to it are load balanced across all healthy nodes. Pass the key in the api-key header, or use the official clients:
curl -X GET https://xyz-example.eu-central.aws.cloud.qdrant.io:6333 \
--header 'api-key: your_api_key_here'
from qdrant_client import QdrantClient
client = QdrantClient(
url="https://xyz-example.eu-central.aws.cloud.qdrant.io:6333",
api_key="your_api_key_here",
)
import { QdrantClient } from "@qdrant/js-client-rest";
const client = new QdrantClient({
url: "https://xyz-example.eu-central.aws.cloud.qdrant.io",
port: 6333,
apiKey: "your_api_key_here",
});
use qdrant_client::Qdrant;
let client = Qdrant::from_url("https://xyz-example.eu-central.aws.cloud.qdrant.io:6334")
.api_key("your_api_key_here")
.build()?;
import io.qdrant.client.QdrantClient;
import io.qdrant.client.QdrantGrpcClient;
QdrantClient client = new QdrantClient(
QdrantGrpcClient.newBuilder("xyz-example.eu-central.aws.cloud.qdrant.io", 6334, true)
.withApiKey("your_api_key_here")
.build());
using Qdrant.Client;
var client = new QdrantClient(
host: "xyz-example.eu-central.aws.cloud.qdrant.io",
port: 6334,
https: true,
apiKey: "your_api_key_here");
import (
"github.com/qdrant/go-client/qdrant"
)
client, err := qdrant.NewClient(&qdrant.Config{
Host: "xyz-example.eu-central.aws.cloud.qdrant.io",
Port: 6334,
APIKey: "your_api_key_here",
UseTLS: true,
})
A successful request returns the server title, version, and commit. Granular access control for keys requires cluster version v1.11.0 or later. See Authentication.
Load Data and Search
- Create a collection, upload points, and run a query with the Cloud Quickstart.
- Generate embeddings inside the cluster with Cloud Inference. It is enabled by default on clusters created after July 7, 2025, and includes free dense, sparse, and multivector models.
- Bring data from another system with the migration guides.
Prepare for Production
A cluster is production-ready when it meets these minimums. The Cloud Production Checklist covers every setting in full.
| Area | Minimum | Where to set it |
|---|---|---|
| Availability | At least 3 nodes, and a replication factor of at least 2 on every collection | High Availability option at cluster creation, and the default replication factor under Collection Defaults |
| Scaling headroom | At least 2 shards per node in each collection | Shard count at collection creation. Resharding changes it later |
| Recovery | A backup schedule | Backups tab of the cluster. See Backups |
| Access | One key per application with an expiration, and the narrowest permission each needs | API Keys tab. Restrict client networks with IP ranges |
| Disk speed (AWS) | Cost-Optimized tier (the default), unless the workload is write-heavy or has a low RAM cache ratio. Then Balanced for more IOPS and throughput, or Performance for the most | Cluster creation, or later |
| Monitoring | Alert emails reaching the account owner, and metrics scraped from /metrics and /sys_metrics | See Monitoring |
| Upgrades | Take a backup, update the cluster, then update the client SDKs. Keep each SDK within one minor version of the cluster | See Updating Clusters |
Multi-AZ clusters, available on the Premium tier, can only be chosen at creation. They need a multiple of 3 nodes. For the reasoning behind each setting, see Vector Search in Production.
Automate
| Tool | Use it for | Start here |
|---|---|---|
| Cloud API | Provisioning from services and scripts, at https://api.cloud.qdrant.io (REST) or grpc.cloud.qdrant.io:443 (gRPC), with a Cloud Management Key | Cloud API |
qcloud CLI | Terminal and CI workflows | Cloud CLI |
| Terraform and Pulumi | Version-controlled infrastructure | Infrastructure Tools |
For the same workflow in each tool, see Automate Qdrant Cloud.
Next Steps
| To do this | Read |
|---|---|
| Invite teammates and set roles | Cloud RBAC |
| Change cluster size | Scale Clusters |
| Tune defaults and restart behavior | Configure Clusters |
| Review compliance and network isolation | Cloud Security |
| Compare Standard and Premium | Premium Tier |