ManyVector on AWS S3.
Store your HNSW index, inverted index, and raw vectors in your own S3 bucket. Standard or Express One Zone storage class. Any AWS region. ManyVector never stores your data — it reads and writes only to your bucket using credentials you provide.
Storage classes
Standard and Express One Zone
Use S3 Standard for multi-AZ durability or S3 Express One Zone for single-digit-millisecond latency at hot namespace prices up to 10x cheaper.
Regions
Any AWS region
Deploy ManyVector query nodes co-located with your S3 bucket to minimize latency. us-east-1, eu-west-1, ap-southeast-1 — any region S3 supports.
Auth
IAM roles and access keys
Provide an IAM role ARN or access key pair scoped to your namespace bucket. ManyVector uses these to read and write objects — never storing credentials on our side.
Encryption
SSE-S3, SSE-KMS, and DSSE-KMS
All objects at rest are encrypted by S3. Use your own KMS key (SSE-KMS) for additional control, or dual-layer encryption (DSSE-KMS) for regulated workloads.
Access logs
Full S3 access logging
Enable S3 server access logging to capture every read and write ManyVector performs on your bucket — for compliance, audit, and anomaly detection.
Cost
$0.023/GB/month (S3 Standard)
You pay AWS directly for storage — ManyVector adds no markup. Typical HNSW index overhead is 1.2-1.5x the raw vector bytes. S3 pricing is fully transparent.
The world's most widely deployed object store.
$0.023/GB
S3 Standard / month
11 9s
durability guarantee
Any
AWS region
SSE-KMS
encryption options
Common questions
Which S3 storage classes does ManyVector support?
S3 Standard (multi-AZ durability) and S3 Express One Zone (single-digit-ms, same-AZ). Standard is the default; use Express One Zone for the lowest-latency hot namespaces.
How do I grant ManyVector access to my bucket?
Provide an IAM role ARN or access key pair scoped to your namespace bucket. ManyVector assumes the role or uses the key to read/write objects. IAM roles are recommended over long-lived keys.
Does ManyVector store data outside my S3 bucket?
No. ManyVector writes all HNSW graph files, inverted index objects, and raw vectors directly to your bucket. The only state outside your bucket is the namespace metadata record in ManyVector's control plane.
Can I use a VPC endpoint to avoid the public internet?
Yes. Configure an S3 Gateway or Interface VPC endpoint and deploy ManyVector query nodes in your VPC. All traffic stays within your network.