Membase-MCP Server is the interface to the Membase decentralized memory layer. It enables AI agents to securely store and retrieve conversation history, interaction records, and knowledge across a distributed Unibase DA network. By conforming to the Membase protocol, the server provides decentralized, verifiable storage that preserves agent continuity, personalization, and traceability. It exposes functions to manage memory contexts per conversation: fetch the current conversation id, switch to another conversation, save a message or memory into the active conversation, and retrieve the most recent messages. This setup allows agents to upload memories and later reconstruct or reference prior interactions, supporting long-term reasoning and user-specific behavior. The server is designed to be integrated via MCP clients such as Claude, Windsurf, Cursor, and Cline and uses environment variables to identify the account, conversation, and instance. It supports memory operations for upload and retrieval and is intended to run as a standalone service that can be deployed in testnet environments, with a testnet endpoint available for memory interactions.