Loading article…
Walrus introduces MemWal SDK and Walrus Memory, a decentralized, encrypted memory layer for AI agents that offers portability, verifiability, and multi‑agent
Walrus, the decentralized storage protocol built on the Sui blockchain, has unveiled a new memory layer for AI agents that stores encrypted data on its network and adds semantic search capabilities, allowing agents to retrieve context by meaning rather than exact keywords [1].
Key takeaways
Walrus’s MemWal SDK is designed to give AI agents a persistent memory that lives on a verifiable data layer rather than on a single provider’s servers. The Sui blockchain handles ownership and access control, meaning users decide who can read, write or share an agent’s memories [1]. The SDK stores encrypted memories on the Walrus network and layers a semantic‑search retrieval system on top, enabling agents to query their own memory intelligently based on meaning [1].
The product is positioned as a solution to what Mystic Labs co‑founder Kostas Chalkias calls the “real bottleneck” in AI—agentic memory. Chalkias argues that current workflows force developers to stitch together databases, vector stores and runtime state, leading to unreliable agents that forget context [3]. Walrus Memory aims to address this by offering four pillars: verifiability, availability, portability and shareability. Verifiability is provided through cryptographic tools such as zk‑proofs that let agents confirm data integrity; availability is ensured by the decentralized network; portability allows memories to move between models and vendors; and shareability enables multiple agents to collaborate using shared memory pools [1][4].
The launch includes direct support for leading AI models—Claude, ChatGPT and Gemini—and plugins for the OpenClaw and NemoClaw agentic frameworks, as well as Python and TypeScript SDKs [3][4]. Additional integrations with the Vercel AI SDK and quick‑start guides are part of the beta release, and the team is gathering developer feedback via GitHub [1]. Early adopters such as Allium, Conso Labs, Inflectiv, OpenGradient, Talus Labs and Tatum are already experimenting with the platform to build portable agent identity systems and assistants that retain customer interactions across sessions [3].
Coverage is mostly measured — 4 of 4 reports stay neutral.
Every Monday — the token unlocks, Fed dates & catalysts set to move crypto and markets this week. So you’re never blindsided.
Free · 3-min read · one-click unsubscribe
OpenGradient is a decentralized infrastructure network designed to host, infer, and verify AI models at scale using specialized GPU and Trusted Execution Environment nodes.
OPG began trading on HTX on April 21, 2026, with both spot and isolated margin pairs against USDT.
The OPG token powers compute and settlement of verifiable inference, rewards model development, enables staking for network security, grants premium access to ecosystem apps, and provides governance over the protocol.
Walrus Memory represents one of the first attempts to provide a blockchain‑backed, portable memory layer specifically for AI agents. By moving memory off proprietary platforms and into a decentralized, encrypted store, the solution promises greater user control, cross‑model compatibility and improved retrieval quality. If the claimed 60 % improvement in ranking and filtering holds in broader deployments, developers could see faster, more reliable agent workflows and reduced engineering overhead for custom storage solutions. The beta launch and growing ecosystem suggest that the next few months will be critical for testing scalability, enterprise readiness, and the ability to meet evolving privacy and compliance requirements.
AI-assisted synthesis by the TrendWatcher Editorial Desk · sourced from 4 outlets · Jun 13, 2026 · How we report
OpenGradient is listed as a partner using Walrus Memory, which provides portable, encrypted, and verifiable memory for AI agents across multiple platforms.