
from memwal12
Privacy-first AI memory SDK: encrypted persistent memories stored on Walrus with on-chain ownership on Sui, plus semantic recall and Vercel AI integration.
MemWal is a TypeScript SDK that provides persistent, encrypted AI memory storage using Walrus for decentralized blobs, SEAL for encryption, and Sui smart contracts for ownership. It exposes APIs to remember, analyze, recall, and restore memories with semantic vector search.
Use MemWal when an app or agent requires end-to-end encrypted persistent memories across sessions or devices, semantic recall by meaning, decentralized storage resilience, or cryptographically enforced ownership and delegation. Not suitable for ephemeral session-only context or large binary files.
Best for agents integrating persistent memory into conversations or applications (agents that can call Node/TypeScript SDKs or use the Vercel AI middleware).
MemWal is a privacy-first AI memory SDK using Walrus decentralized storage with SEAL encryption and Sui on-chain ownership. The SKILL.md is exceptionally thorough — covering API surface, configuration, Vercel AI SDK integration, OpenClaw plugin setup, and detailed troubleshooting. Scripts are TypeScript/ESM integration tests and a credential verifier; none could run due to the runner not supporting .ts/.mjs extensions. One script (test-zero-restore.ts) contains a hardcoded Ed25519 private key, which is a security concern even if intended for testing.
node/tsx (scripts are .ts/.mjs, runner only supports .py/.sh/.js)Well-structured skill from MystenLabs (Sui/Walrus team). The hardcoded private key in test-zero-restore.ts is the main security deduction — likely a throwaway test key but sets a bad pattern. SKILL.md documentation quality is among the best audited: complete API reference, response shapes, namespace semantics, and integration guides.