Take Your Agent’s Memory Anywhere: Introducing Walrus Memory

Memory: AI’s next big bottleneck

Ask any developer building production AI agents, and the limiting factor is increasingly becoming memory. Agents lose context between sessions. Workflows restart from zero. Multi-agent systems hit the limits of context windows. Builders are left to stitch together databases, vector stores, and runtime state, resulting in unreliable systems that buckle under longer and more complex workflows.

A new kind of memory infrastructure

Walrus Memory is the first memory layer built specifically for AI agents that is portable, verifiable, and fully under builders’ control. This lets agents carry context across apps and sessions, share memory with other agents, and verify the data they act on.

• Portable by design. Memory moves freely across agents, apps, and workflows. Not tied to a single runtime, session, or provider.
• Fully under your control. Memories are encrypted by default, and owned by you. Programmable access controls allow builders to define exactly who can read, write, and share memory. Delegate access to agents and workflows, and revoke at any time.
• Built for agent coordination. Shared memory spaces let multiple agents stay aligned across long-running workflows. Every memory carries proof of its integrity, so agents can verify the context they act on.

Plugs into what you already use

Walrus Memory integrates with leading AI platforms including Claude, ChatGPT, and Gemini. With plugins for OpenClaw and NemoClaw, native MCP support, and SDKs for Python and TypeScript, developers can add portable memory to existing agent workflows in just a few lines of code.

“Memory is one of the most critical bottlenecks in AI today,” said Kostas Chalkias, Co-Founder and Chief Cryptographer at Mysten Labs, the original contributor to Walrus. “Most agent memory lives locked inside platforms. Walrus Memory changes this. It puts builders in control and lets agents move and collaborate across different services. This is such an important foundation for the agentic future we all see coming."

What teams are building

Teams integrating Walrus Memory today include Allium, Conso Labs, Inflectiv, OpenGradient, Talus Labs and Tatum – exploring use cases from portable agent identity systems and to AI assistants that remember customer interactions across sessions.

"Portable memory across AI systems is a huge unlock. Engineers already bounce between OpenAI, Anthropic, and Gemini, and switching between platforms means rebuilding context from scratch. Walrus Memory is helping make persistent, portable context a foundational piece of AI infrastructure." – Ethan Chan, Co-Founder and CEO, Allium

“Walrus Memory is going to let our monitoring agents retain prior observations, so they can avoid reprocessing the same events. Every decision will also produce an audit trail, allowing us to verify and explain why activity was flagged.” – Dion Cornett, CEO, Tatum

“As AI systems run on their own and for longer stretches, their memory infrastructure calls for stronger guarantees around verifiability, portability, and reliability. We see Walrus Memory as part of a major shift toward open and interoperable AI systems.” – Advait Jayant, Chief Strategy Officer, OpenGradient

Built for the next generation of AI agents

In a future where millions of AI agents are coordinating and transacting on our behalf, the memory layer they run on needs to hold up at scale. Walrus Memory is built on Walrus, infrastructure engineered for exactly that: verifiable data, the performance to scale, and access controls that hold up across organizations.

Take your agent’s memory anywhere.

Get started for free at walrus.xyz/memory