Matterhorn and Walrus Bring Persistent Memory to AI-Assisted Blockchain Workflows
Matterhorn is storing its AI agents' project memory on Walrus, so developers building blockchain apps in natural language keep their context across sessions and teams.
Matterhorn uses Walrus as a memory layer to preserve context across AI-assisted blockchain workflows. Users can interact with blockchain applications and create custom workflows through natural language while retaining control of transaction approval and signing.
Matterhorn is an AI coworker that makes Web3 easier to use through chat. Users can interact with protocols and prepare transactions while retaining control of approval and signing through their own wallets. Matterhorn uses Walrus Memory for agent memory and Artificial Superintelligence (ASI) Alliance’s models to support its AI workflows.
Key takeaways
-
Matterhorn is making Walrus Memory the default backend for agent memory on its platform. Project context, deployment history, audit results, and agent conversations persist across sessions and are shared across a team.
-
Most AI agents retain no memory between sessions. Matterhorn users can interact with protocols using chat, just like they use Claude Cowork or Codex, with Walrus providing the persistent memory underneath.
-
Memory stored on Walrus is tamper-evident, enabling users and developers to verify an agent's record of what it did, and what it built.
Agents without memory: a key constraint for builders
Most AI agents retain no memory between sessions. Project context, audit findings, and deployment history disappear when a session ends and have to be rebuilt before work can resume. Matterhorn is solving for this by combining a chat interface with persistent, verifiable memory stored on Walrus, so users can continue their session without repeatedly explaining the same background.
Project memory that persists
On Matterhorn, project context is now stored on Walrus through Walrus Memory, the portable memory layer for AI agents. This includes everything an agent learns about a project, from documentation to deployment history and audit reports.
The first advantage for users is persistence across a project. Memory written by one agent is available to every agent on the project and every member of the team, in every subsequent session. Context is built once and reused, which means less rework and more useful assistance the longer a project runs.
The second is verifiability. On Walrus, any party can independently prove that an agent's memory has not been altered since it was stored, and that it is still available. When an agent is executing transactions with real consequences at stake, that means the integrity of the data informing those decisions can be proven.
“Over seven years building crypto products, I have repeatedly seen how difficult even simple tasks can be for users,” said Abhinav Ramesh, Founder and CEO of Matterhorn. “We are building Matterhorn to make Web3 usable through chat while keeping users in control of their funds. Walrus gives us a decentralized memory layer that helps preserve context across those workflows.”
Built for systems of consequence
"These advantages also matter for developers building products on Matterhorn. Shared memory across agents and team members makes workflows more efficient, and audit findings and deployment records can be proven to third parties rather than taken on trust. This becomes more critical as blockchain moves into high-stakes enterprise settings, including tokenization and payments, and as workflows become more complex and multi-step."
"AI is now the primary way software gets built, and the apps it builds involve significant money and consequences," said Kostas Chalkias, Co-Founder and Chief Cryptographer at Mysten Labs, original contributor to Walrus. "Systems like this need a bedrock of memory that persists and data that anyone can verify. That is what Walrus excels at."
Live today
Matterhorn Web and Matterhorn Pro support custom workflows across more than 20 blockchains. Matterhorn Desks brings a chat-first experience to Web3 workflows. Through the ASI Alliance and Walrus partnership, more than 5,000 users in the community can build on Matterhorn using decentralized AI models and infrastructure, with Walrus as storage for the apps they build.
Explore Matterhorn at matterhorn.so, and visit walrus.xyz to learn how Walrus stores data that anyone can verify.
FAQs
What is Matterhorn?
Matterhorn is an AI coworker that makes crypto easier to use through chat. Describe what you want to do and it helps you use Web3 products and prepare transactions. You review and approve transactions in your own wallet. Matterhorn never holds your funds or signs on your behalf. Matterhorn uses Walrus as its memory layer to store context for future interactions.
What is Matterhorn storing on Walrus?
Matterhorn stores the persistent memory layer for its AI agents on Walrus, including project context, AI conversations, technical documentation, deployment history, smart contract artifacts, audit reports, and workflow state. This is the structured knowledge an agent needs to maintain context across development sessions, and that teams need to share and verify.
Can I verify that an agent's memory hasn't been changed?
Yes. Memory stored through Walrus is tamper-evident, so any third party can independently confirm that a record hasn't been altered since it was written. This matters most for workflows with consequences, like transaction history or audit findings, where trust in the underlying data is essential.
How does Matterhorn give AI agents memory that persists across sessions?
Matterhorn uses Walrus Memory as the memory layer behind its AI agents, so project context, conversation history, and workflow state are written to Walrus rather than kept only in a chat session. That means context carries over the next time a user or agent picks up a project, instead of resetting to zero each session.
About Walrus
Walrus is the data platform built for the demands of AI. It lets builders, enterprises, and AI agents move data across apps and models, protect it with built-in access rules, and make it verifiable by anyone. As AI takes actions with consequence, the data it runs on has to be provable. Created by the cryptographers and distributed systems engineers behind Sui and Meta's first stablecoin initiative.
About Matterhorn
Matterhorn is building an AI coworker that makes Web3 usable through plain language. Founded by a team with seven years of experience building crypto products for businesses and consumers, Matterhorn helps users navigate protocols and prepare transactions while keeping approval and signing in their own wallets. It uses Walrus for its memory layer and ASI Alliance’s MeTTa programming language and AI models to support its AI workflows.
