Walrus Launches MemWal SDK for AI Agent Memory

Walrus has launched MemWal, an SDK designed to address limitations in agentic memory by bringing verifiability, availability, portability and sharability to how AI agents store and access information, according to Mysten Labs Group Product Manager Abinhav Garg.

Verifiable and Portable Memory Layer

MemWal stores memory on an open, verifiable data layer that is not tied to any single model or vendor. This enables users to switch between different model providers such as OpenAI and Anthropic while maintaining data with verifiable guarantees that make it tamper-proof. “With Walrus plus MemWal, memory lives on an open, verifiable data layer, so that means it’s not tied to any one model or vendor,” Garg explained to Decrypt.

Data stored on Walrus inherits built-in guarantees around verifiability, portability and availability, enabling “easier sharing of memory between agents across teams and organizations,” Garg said, describing this capability as “a must for agent collaboration.”

Framework Integration

MemWal integrates with popular agent orchestration frameworks OpenClaw and NemoClaw through a plugin released this week. The integration is designed to simplify adoption by allowing developers to equip their agents with durable, verifiable memory using tools they already work with. “Without this, developers would have to understand the integration of a decentralized storage layer like Walrus, which could add friction and complexity,” Garg explained.

Privacy and Access Control

MemWal includes privacy features through a native encryption layer and programmable access control. Even though the storage itself is decentralized, the contents remain confidential and governed by policy—“even the storage providers cannot read it,” Garg stated. This approach addresses growing concerns about agents handling sensitive and proprietary data, including enterprise workflows, financial information and personal context.

Emerging Use Cases

The enhanced agentic memory capabilities enable new applications across multiple domains. Customer support agents can retain contextual cues about users, and agents across different teams can collaborate by “working off the same customer history.” Other partners are exploring coordination between agents operating as publishers or consumers on marketplaces, using messaging as a form of shared memory. Additional use cases include robots that need to share context with each other to coordinate tasks in real-world scenarios, such as disaster response operations.

Garg anticipates a “standardization of the stack” for agents in the future, with “clear separation between compute, data, memory and coordination.” “Our view is that memory and data shouldn’t be tied to any single model or platform—so Walrus becomes that durable data layer and MemWal becomes a memory layer on top of it,” he said.