Unibase, also known by its token symbol UB, is a crypto/Web3 project focused on creating a high-performance decentralized AI memory layer. Its main purpose is to equip autonomous AI agents with long-term memory capabilities, improve their interoperability, and ensure data sovereignty. This initiative aims to address existing gaps in the AI ecosystem, specifically concerning persistent memory, communication between various AI agents, and user control over their data. The project is designed to lay the groundwork for what it terms the "Open Agent Internet." Unibase achieves its goals through three core modules. The first, Membase, offers secure and scalable storage specifically for the long-term memory of AI agents, incorporating zero-knowledge proof-based validation for memory verification. The second module is the AIP Protocol, a Web3-native standard facilitating communication, identity management, and shared states among different agents. The third component, Unibase DA, is a data availability layer that uses zero-knowledge verification to provide high throughput, exceeding 100 gigabytes per second, to support real-time AI processing. These integrated components enable AI agents to retain information, collaborate effectively, and evolve across different platforms, all while ensuring that data access is verified and decentralized. The UB token plays a crucial role within this ecosystem. It is used to cover protocol fees for services like memory storage, agent deployment, and interoperability. Additionally, the token is involved in governance through a veUB model, agent staking, and incentives for knowledge mining. Unibase is planned for deployment on multiple blockchain networks, beginning with Ethereum and BNB Chain, with future expansions to other Ethereum Layer 2 networks. The ecosystem built around Unibase includes projects such as BitAgent for multi-agent collaboration, TradingFlow for autonomous trading agents, TwinX for self-evolving agents, and Beeper for intent-based Twitter agents. It also integrates with various frameworks, including MCP, Virtuals, ElizaOS, and Swarms. The project's testnet demonstrated early adoption by mid-2025, recording over 200 deployed agents and more than 12.4 million on-chain memory entries. Unibase distinguishes itself by focusing on the infrastructure for persistent memory, verifiable execution, and inter-agent collaboration, rather than on AI models or applications alone.
Learn more