AIOZ Storage Developer Platform with SDKs, CLI, and MCP

The AIOZ Storage developer platform is the set of tools you can build with: the standard AWS SDKs and CLI for buckets and objects, the AWS S3 Vectors SDKs for the Vector Database, Super Intelligence Memory SDKs, and an MCP server. This hub maps each one to its guide.

TL;DR:

• For buckets and objects you use the AWS SDKs for JavaScript, Go, and Python, the AWS CLI, or signed requests in Postman. rclone, Terraform, and two application tutorials build on the same setup.
• For vectors you use the AWS S3 Vectors SDKs: boto3's s3vectors client, @aws-sdk/client-s3vectors, or Go's s3vectors package.
• Super Intelligence Memory has its own SDKs for Go, JavaScript/TypeScript, and Python, and the MCP server exposes 22 S3 and vector tools to MCP-compatible assistants.
• The S3 tools share one setup: the endpoint https://s3.aiozstorage.network, path-style addressing, region us-east-1, and credentials from an access grant.

Updated Oct 19, 2026: added the Vector Database SDKs, the Super Intelligence Memory SDKs, and the MCP server from the current docs, and corrected the region note.

What setup do all the S3 tools share?

AIOZ Storage is S3-compatible, so the AWS SDKs, the AWS CLI, Terraform's AWS provider, Postman, and rclone already know how to talk to it. Each one needs the same four changes. You point it at https://s3.aiozstorage.network instead of AWS, turn on path-style addressing, and set the region to us-east-1, the value the docs' examples use. Then you give it credentials from an AIOZ Storage access grant instead of an AWS IAM key.

Every object-storage guide below is that same swap, written in the configuration format each tool expects. Once you've done it in one tool, the rest are quick.

Which tool should you start with?

Writing application code? Use one of the three language paths: JavaScript via @aws-sdk/client-s3, Go via aws-sdk-go-v2, or Python via boto3. All three are official AWS SDKs pointed at AIOZ's endpoint, so there is no separate client library to learn for object storage.

Scripting or one-off operations? The CLI quickstart points the standard AWS CLI at AIOZ Storage's endpoint. Use it for anything you'd normally do with aws s3, without writing a script around an SDK first.

Managing buckets as infrastructure? Terraform manages AIOZ Storage buckets with the same aws_s3_bucket resource and state file you use for AWS. The AIOZ docs don't cover Terraform, so that guide is our own setup.

Need something the SDKs and CLI hide? Calling the API directly with Postman shows how to build AWS Signature Version 4 requests by hand, so you can see and control every header.

Vector Database SDKs

The docs say "AIOZ Storage works with the AWS S3 Vectors SDKs," with step-by-step guides for Python, JavaScript, and Go. In Python you install boto3 and create a client for the s3vectors service. In JavaScript you install @aws-sdk/client-s3vectors, and in Go you add github.com/aws/aws-sdk-go-v2/service/s3vectors.

The flow is the same in each language. You create a vector bucket, create an index with a dimension that matches your embedding model, put vectors with optional metadata, and query for the nearest matches. The credentials come from the dashboard, the region is us-east-1, and the SDK pages leave the endpoint as a placeholder for you to fill in. Vector usage is billed separately from object storage, per the billing page.

Super Intelligence Memory SDKs

Super Intelligence Memory gives an agent long-term memory with two operations: Record saves a plain-language statement and Recall returns the saved statements closest to a question. You don't create embeddings or indexes yourself. The official SDKs are super-intelligence-memory on pip (Python 3.10 or later), @aiozstorage/super-intelligence-memory on npm, and github.com/AIOZStorage/super-intelligence-memory-go (Go 1.25 or later).

Two setup steps catch people out. The bucket must be created once with memoryManaged set to true, through a signed request, because the SDKs can't create it. And the base URL is https://vector-api.aiozstorage.network, with no trailing slash, or the signature check fails.

The AIOZ Storage MCP server

If you'd rather have an assistant do the work, the MCP server lets any MCP-compatible client manage storage and run vector search through natural-language prompts. It connects over HTTP SSE at https://mcp.aiozstorage.network/sse and documents 22 tools: 3 for S3 buckets, 6 for S3 objects, 4 for vector buckets, 4 for indexes, and 5 for vectors.

Every tool call carries S3_ACCESS_KEY_ID and S3_SECRET_ACCESS_KEY in its arguments. S3 tools need S3 Access Grant credentials, and vector bucket tools need a Vector Access Grant with All Buckets scope. Give the assistant keys with only the permissions the task needs.

How do you move existing data in?

Two migration guides cover the common starting points, and both use rclone. Migrating from Amazon S3 is the simpler one: source and destination both use rclone's s3 remote type, so you change provider = AWS to provider = Other and add the new endpoint and credentials.

Migrating from Backblaze B2 looks similar but isn't. B2's native remote type in rclone (type = b2) predates its S3-compatible API, so that move goes from B2-native to S3-compatible instead of staying on one protocol.

Building a real application

Two tutorials go past single-operation quickstarts. The FastAPI file-management API wires boto3 into a working upload and download HTTP API, the shape most backend services need. Generating presigned URLs shows how to hand out temporary, scoped access, so a browser or mobile app can upload straight to a bucket without holding your credentials.

Automating what runs on a schedule

Two more guides cover unattended jobs. Automating backups with GitHub Actions runs the rclone remote from the migration guides on a schedule. Multi-cloud disaster recovery with rclone bisync keeps AIOZ Storage and AWS S3 mirrored in both directions. Both reuse the same rclone remote, run repeatedly instead of once.

Why build on existing S3 tooling

Because AIOZ Storage uses existing S3 and S3 Vectors tooling, your team's AWS skills carry over. rclone's documentation says its S3 backend "can be used with a number of different providers," and it keeps "Other" as a provider option for "any other S3 compatible provider." If you already have rclone remotes for another cloud, AIOZ Storage is one more endpoint, and the SDK guides tell the same story for application code.

How AIOZ Storage handles a first request

The quickest test of your credentials is the AWS CLI. The last two lines come from the docs' CLI page: one creates a bucket and the other lists your buckets.

export AWS_REQUEST_CHECKSUM_CALCULATION=when_required
aws configure set default.s3.addressing_style path
aws s3 --endpoint-url https://s3.aiozstorage.network mb s3://my-bucket
aws s3 --endpoint-url https://s3.aiozstorage.network ls

The first two lines are ours, not from the docs' CLI page. mb needs path-style URLs (AWS CLI S3 configuration) because bucket subdomains don't resolve. The when_required line matters once you upload: AIOZ Storage rejects uploads without a Content-Length header, which AWS CLI v2.23.0 and later send by default (AWS CLI environment variables). We confirmed both on Oct 19, 2026; a full upload with real keys is not tested yet.

For vectors, the docs' Python guide starts with pip install boto3 and a boto3.client('s3vectors', region_name='us-east-1', ...) call. What the docs don't cover yet: Terraform, a published vector endpoint on the SDK pages, and MCP client setup beyond the SSE URL.

Next step: pick your language in the AIOZ Storage SDKs and CLIs docs.

Frequently Asked Questions

What's the fastest way to start building on AIOZ Storage?
It depends on what you're building: a language SDK for application code, the CLI for scripting, Terraform for buckets as infrastructure, or the S3 Vectors SDKs for embeddings. The S3 tools all share one endpoint-and-credentials setup.

Does AIOZ Storage have its own SDK?
For object storage and the Vector Database, no: you use the official AWS SDKs and AWS S3 Vectors SDKs pointed at AIOZ. Super Intelligence Memory is the exception, with AIOZ's own SDKs for Go, JavaScript/TypeScript, and Python.

Which SDKs work with the AIOZ Storage Vector Database?
The AWS S3 Vectors SDKs. The docs have guides for Python (boto3), JavaScript (@aws-sdk/client-s3vectors), and Go (aws-sdk-go-v2/service/s3vectors).

Does AIOZ Storage have an MCP server?
Yes. It runs over HTTP SSE at https://mcp.aiozstorage.network/sse and documents 22 tools for S3 buckets, objects, vector buckets, indexes, and vectors. Each tool call carries your access keys.

How do I move existing data from AWS S3 or Backblaze B2 to AIOZ Storage?
With rclone. The S3 migration is a config change (provider = AWS to provider = Other). The B2 migration is a bigger jump, from B2's native remote type to an S3-compatible one.

Can I manage AIOZ Storage buckets with Terraform?
Yes, through the standard Terraform AWS provider with a custom S3 endpoint, s3_use_path_style = true, and a few validation-skipping flags. The AIOZ docs don't cover Terraform.

What if I need something the SDKs and CLI don't support directly?
Call the API directly. The Postman guide covers building and signing raw S3-compatible requests with AWS Signature Version 4.

Can I automate recurring backups or keep two clouds in sync?
Yes. A scheduled GitHub Actions workflow handles one-way backups, and rclone bisync keeps AIOZ Storage and AWS S3 mirrored in both directions.

Related Reading

• What Is AIOZ Storage? S3-Compatible Decentralized Storage - the full product model every guide on this hub builds on top of.
• AIOZ Storage Access Control: Macaroons and Access Grants - generating the credentials every tool on this hub needs.
• How to Use AIOZ Storage Effectively: A Practical Checklist - the setup mistakes to avoid before wiring any of these tools up.
• The 3-2-1 Backup Rule: A Practical Framework Explained - the framework behind the backup and disaster-recovery guides on this hub.
• How Automatic MIME Type Detection Works on AIOZ Storage - a real upload-time behavior that applies across every SDK, CLI, and API path on this hub.
• What Is CORS and Why It Matters for Object Storage? - a real requirement for any browser-based upload path built on top of the SDKs and APIs covered here.
• What Is an MCP Server? How Agents Access Storage - the protocol behind AIOZ Storage's MCP server, which covers S3 and vector tools.
• What Is Multipart Upload? Splitting Files Into Chunks - how the SDKs and CLI on this hub build large objects behind the scenes.
• AIOZ Storage llms.txt: Docs Built for Super Intelligence - the machine-readable docs files AIOZ publishes for coding assistants and retrieval tools.
• AIOZ Storage Super Intelligence Memory: Record and Recall - the Go, JavaScript, and Python SDKs for agent memory, step by step.
• Vector Database vs Agent Memory on AIOZ Storage: A Guide - which vector path to build on.

References

• AIOZ Storage Documentation: CLI
• AIOZ Storage Documentation: Vector Database
• AIOZ Storage Documentation: Vector Database Python SDK
• AIOZ Storage Documentation: Super Intelligence Memory
• AIOZ Storage Documentation: MCP Server
• rclone Documentation: Amazon S3 (and S3-compatible) Storage Providers
• AWS CLI Documentation: Using Endpoints