AIOZ Storage MCP Server Review: AI Agent Tools, S3 and Vector Capabilities, Access and Pricing

The AIOZ Storage MCP Server launched on October 12, 2026, and lets MCP-compatible AI agents execute storage operations through natural language prompts. It provides 22 specialized tools that handle both S3-compatible object storage and vector database tasks without requiring separate code for each action.

Core Capabilities and How It Works

The server runs at the endpoint https://mcp.aiozstorage.network/sse using HTTP SSE transport. Agents authenticate each tool call with an AIOZ Storage access key ID and secret access key; no separate login step exists. Supported clients include VS Code, Cursor, and Claude. The 22 tools split into nine S3 operations and 13 vector operations.

S3 tools cover bucket listing, creation, and deletion plus object listing, metadata retrieval, upload via presigned URL, copying, and deletion. The put_s3_object tool returns a presigned URL that expires after 900 seconds by default. The get_s3_object tool caps responses at roughly 1 MiB; larger files require a direct S3 client. Vector tools manage vector buckets and indexes up to 4,096 dimensions with cosine, Euclidean, or dot-product distance metrics. Agents can insert up to 500 vectors per call with metadata, retrieve or delete them, and run filtered queries that return between one and 100 results.

Access requires either an S3 Access Grant or a Vector Access Grant created through the AIOZ Storage dashboard. Grants can be scoped to all buckets or a single bucket. The MCP server itself carries no published usage fees beyond the underlying storage costs.

Pricing Structure and Security Measures

Object storage costs approximately $17 per TB per month plus $5 per TB for delivery and is billed in AIOZ tokens on a pay-as-you-go basis with no per-request charges. Vector storage is priced at $0.05 per GB-month. Additional vector fees include $0.04 per million tokens for embeddings, $2 per million queries, and separate processing and return charges. These rates apply to the storage layer; the MCP server layer adds no extra published cost.

Each tool call transmits credentials inline, so clients should store keys securely and avoid logging them. Grants support fine-grained permissions that limit exposure. The service forms part of AIOZ Network’s decentralized storage ecosystem and follows the open MCP standard for tool integration.

Early users gain direct agent-driven bucket and index management without writing custom API code. The system also prepares for upcoming AIOZ Agent Memory features that rely on the same vector infrastructure.

Limitations, Competitor Landscape, and Verdict

Rate limits specific to the MCP server remain unspecified in current documentation. Vector query results are capped at 100 items per call, and large object retrievals require fallback to standard S3 clients. No public adoption metrics or independent performance benchmarks have been released since the October 12 launch.

Compared with MinIO’s preview MCP offering, AIOZ provides a larger published tool set and explicit vector pricing. Against AWS S3 combined with Bedrock agents, AIOZ charges lower base storage rates but requires payment in its native token and operates on a newer decentralized network. Both alternatives currently demand more manual integration work for equivalent agent actions.

The service remains in an early post-launch phase with limited public user feedback available. No widespread complaints or confirmed incidents have surfaced, yet the absence of independent audits and rate-limit details leaves some operational questions open. Developers testing agent workflows will find concrete tool coverage for common storage tasks but should monitor official documentation for updates on limits and new features.