In a significant step towards decentralizing AI memory management, Walrus has launched the MemWal SDK, providing a framework for AI agents to maintain secure, persistent memories. This toolkit empowers users by making memory storage verifiable and portable, allowing agents to recall information based on semantic relevance rather than simple keyword matching.
The Functionality of MemWal SDK
Walrus’s MemWal SDK operates on the Sui blockchain, which serves as the foundation for the decentralized storage protocol. Its core innovation is the ability to store encrypted memories, enhancing security and privacy. By adding semantic search capabilities to this storage, AI agents can retrieve information contextually, enriching their interactions and decision-making. This method surpasses traditional information retrieval techniques that rely solely on exact matches, making AI agents more sophisticated in managing their learned experiences.
Openness and User Control
Abinhav Garg, Group Product Manager at Mysten Labs, highlights the importance of openness in the MemWal ecosystem. The SDK enables memories to exist on a verifiable data layer that does not depend on any single AI provider, ensuring that ownership and access remain with users. The Sui blockchain is essential here, governing who can read, write, or share an agent’s memories and effectively decentralizing control while fostering a collaborative environment.
The implications of this model are significant. Users can now manage their AI memories independently of corporate infrastructures, which often impose terms and conditions that may not align with user interests. This user-centric approach to memory ownership introduces a new paradigm in AI development and deployment.
Integration and Developer Engagement
Walrus is not launching MemWal in isolation. The SDK includes built-in integrations for the Vercel AI SDK and plugins for frameworks like OpenClaw and NemoClaw. To enhance developer engagement, the team has provided quick-start guides and extensive documentation, actively seeking feedback on GitHub as the SDK remains in its beta phase. This proactive approach aims to refine the toolkit based on real-world developer experiences and needs.
Why Decentralized Memory Is Essential for AI
Decentralized memory for AI agents is crucial. The MemWal SDK is based on four fundamental principles: verifiability, availability, portability, and shareability. Verifiability ensures that memories are tamper-proof, which is vital for trust in AI systems. Availability guarantees that memories persist even if a service provider becomes inoperative, protecting user data from unexpected outages. Portability allows memories to be transferred across different AI models and platforms, while shareability promotes collaboration among multiple agents and users, enhancing the collective intelligence of AI systems.
These principles outline a vision for an AI ecosystem where memory management is decentralized, user-friendly, and collaborative, paving the way for more advanced AI applications.
The Future of Decentralized AI Memory
As Walrus continues to develop the MemWal SDK, the groundwork laid since March 2025 indicates a strong commitment to establishing decentralized storage as a foundational element of AI infrastructure. This initiative is likely to attract developers eager to innovate in the AI space while prioritizing user control and collaboration. The evolution of AI agents equipped with user-owned memories could transform AI applications, ultimately leading to more ethical and effective AI systems.
The stories that move AI & crypto markets — before the market reacts.
Free. 7am ET. Five stories. 62,400 readers.
