AI agent sandboxes need external storage for files that outlive the sandbox. Real providers like Modal and E2B mount S3-compatible buckets, AIOZ Storage included.

AIOZ storage can back LangGraph's S3 checkpoint offload tier, but not the whole backend. Here is the real DynamoDB-plus-S3 setup and its credential gap.

AIOZ storage for AI datasets means S3-compatible buckets for training data and model outputs, no native versioning or lifecycle policies. Here is the honest scope.

AIOZ storage for AI workloads means S3-compatible object storage for datasets, checkpoints, and model outputs. No vector database. Here is what is real.

Add AIOZ storage team members through the dashboard's 3-step wizard: name and password, per-bucket permissions, and a one-time credential download.

No Postman collection to import. AIOZ Storage docs show building raw S3 requests by hand, authenticated with AWS Signature and your access grant keys.

No AIOZ-specific CLI exists. Point the standard AWS CLI at --endpoint-url https://s3.aiozstorage.network and every aws s3 command works as documented.

The AIOZ storage go sdk is aws-sdk-go-v2 with a custom endpoint resolver and UsePathStyle set. Here is the exact client setup and upload code.

A FastAPI Boto3 AIOZ Storage upload API is a standard S3 client pointed at a different endpoint and one path-style config flag. Here is the full code.

Backblaze B2 uses rclone's native b2 remote, not s3, so this migrate backblaze b2 storage guide covers a bigger config change than S3, plus egress costs.

AIOZ's own docs formally define only two terms. This AIOZ storage glossary explains bucket, passphrase, macaroon, access grant, eKey, zKey, and sub-user.

AIOZ Storage costs less than AWS S3 on paper, but AWS also charges for requests, storage tiers, and cross-region transfer. Here is the full cost picture.