Abstract
AI agents need memory that survives sessions, vendors, infrastructure shifts, and organizational boundaries. LTM proposes a memory control plane with cryptographic identity, encrypted payloads, verifiable proofs, and governed retrieval.
Whitepaper
Architecture for portable agent memory.
A concise technical model for identity-bound memory, tiered persistence, privacy, token economics, and governance.
Architecture
The system separates agent identity, encrypted memory objects, semantic indexes, proof registries, and policy evaluation. This separation keeps deployments flexible and avoids placing sensitive content on public ledgers.
Privacy
Payloads are encrypted client-side. Enterprises decide retention, region, and storage backend. On-chain data is limited to proofs or non-sensitive metadata because immutable systems cannot honor deletion in the normal web sense.
Roadmap
Foundation SDKs, hosted pilot API, enterprise policy packs, audit trails, private deployment paths, and ecosystem adapters.