Talus Combines Walrus and Sui to Enable Onchain AI Agents
Tokenized AI agents can execute workflows directly onchain, seamlessly accessing data from Walrus while carrying out their functions on Sui.
Talus is a platform for launching onchain AI agents, providing developers with the infrastructure and tooling to build autonomous, income-generating agents. The Talus technology stack includes Talus Network, a network of nodes optimized for agent execution and Nexus, a modular framework for building, composing, and verifying agent workflows.
Through Nexus, Talus enables developers to build composable AI agents that execute workflows directly onchain, enabling secure, transparent, and verifiable AI automation. Talus also plans to launch Idol.fun, a platform that hosts native AI agent experiences, turning Talus infrastructure into real value for end-users through games, interactions and tokenized applications.
Challenge
Talus' agentic framework is built on Sui, and its AI agents need to access large datasets. Although cloud storage is a common solution for this type of system, many of Talus' use cases involve onchain activity, so data would need to be copied from Sui to cloud storage, then retrieved by agents operating on Sui. Centralized cloud storage adds availability risk, creates costs, and includes operational inefficiency.
In addition, more complex agents require higher data demands—from training data to context to memory storage—which makes the ability to easily scale Talus’ data storage a critical requirement for their maturing agentic ecosystem.
Solution
Talus turned to Walrus as its data storage layer to provide highly performant, scalable, and reliable data storage while upholding the company's ethos of decentralization. Walrus stores data used for such purposes as training, context, and memory for Talus' AI agents. The Nexus framework built on Sui serves as a Workflow Engine that orchestrates onchain agent flows, manages state transitions, and coordinates reads/writes to Walrus, Talus’ storage layer, while handling related payments. Developers can also build custom Talus Agent Packages (TAPs), giving them full control over agent logic, external integrations, and data handling, all while leveraging Walrus for scalable, verifiable offchain storage.
The synergy of using Sui for Talus’ coordination and value layer—and Walrus as its decentralized data layer—provides builders on Talus with benefits that outweigh the sum of their parts. Specialized agents can use Walrus to easily store and analyze historic transaction patterns on Sui, unlocking access to a trusted onchain history to audit, formulate strategies, or train new models on onchain behavior. DeFi agents can respond to liquidity needs in real-time, fetching and loading a model instantly from Walrus instead of pulling data from offchain storage. Together, Sui and Walrus bring seamless transparency, security, and reliable data availability to Talus’ agentic workflows.
"Walrus provides the highly performant, secure, scalable and reliable data storage guarantees that are essential for the Cambrian explosion in AI agents we’re about to unlock."
Why Walrus?
Walrus is the only decentralized storage network capable of making large datasets reliably accessible for AI training and usage. As Talus uses Sui as its coordination layer to define and orchestrate agent workflows, using Walrus for data storage was an obvious choice. Talus' AI agents will be able to seamlessly access data from Walrus while carrying out their functions on Sui.
.jpg)