MinIO AIStor Memory pitches enterprise-controlled agent memory
MinIO says AIStor Memory gives AI agents persistent workspaces on customer infrastructure, but pricing and benchmarks were not disclosed.
By Renata Fuchs · Policy Reporter
· 3 min read
MinIO AIStor Memory is the object storage vendor's bid to make long-running AI agents keep their state on infrastructure controlled by enterprise customers. The company says the product extends its AIStor object storage software so agents can retain context across sessions, resume interrupted work and operate on governed enterprise data; MinIO did not disclose pricing, benchmarks or named customers.
The pitch is aimed at a gap between chatbots and agentic systems. A chatbot can answer a single prompt with limited retained context. An AI agent may work through a software task, research assignment or document workflow over hours or days, which creates a record of decisions, intermediate work and learned context that an enterprise may not want scattered across separate systems.
MinIO describes that accumulated output as organizational knowledge. AB Periasamy, MinIO's co-founder and co-CEO, said agent-generated knowledge becomes organizational memory and should remain on enterprise-controlled infrastructure. He said AIStor Memory combines long-term memory, persistent workspaces and secrets on that foundation.
What is AIStor Memory?
AIStor Memory is MinIO's proposed memory layer for AI agents, built into AIStor so memory is treated as a native data type alongside objects and tables. MinIO says agents can access Memory, Workspace and Vault through HTTPS or a POSIX folder mount, which is meant to let teams use current sandboxes, tools and frameworks without changing them.
The company argues that current agent stacks often require teams to assemble object storage, vector databases, metadata stores, secrets managers, governance products and synchronization pipelines. AIStor Memory is positioned as an integrated replacement for that bundle, although MinIO's claims are product claims rather than independently tested performance data.
MinIO says AIStor Memory sits beneath the KV-cache layer served by its MemKV technology. In practical terms, the company is separating short-lived inference context from the larger working state of an agent: files, workspace history, secrets and other information needed to continue a task after an interruption.
Where MinIO says agent memory fits
MinIO listed four target use cases: software engineering agents that work across large codebases, research and analysis jobs that run for extended periods, human-in-the-loop processes that stop and restart, and enterprise AI systems that touch regulated or governed data.
The product inherits storage controls from AIStor, according to MinIO, including erasure coding, bitrot protection, encryption, compression and fault-tolerance features intended to protect against drive, rack and datacenter failures. The company also says memory remains on customer-owned infrastructure and is protected by customer-held keys, rather than leaving the customer's environment.
MinIO further claims AIStor Memory can provide what it calls infinite context, meaning memory scales with available storage capacity rather than the model's context window. That does not make an AI model itself unlimited. It means MinIO is offering a persistent store where data does not have to be removed, summarized or evicted from the memory system when the model's active context is full.
The broader claim is that memory created for one agent can become a shared organizational record built by both employees and AI systems. HYCU and IT Brand Pulse have also surfaced the idea of enterprise memory tied to AI activity, according to MinIO's framing, which suggests the category is moving from chatbot UX into infrastructure buying decisions.
This story draws on original reporting from The Register.
