Cross-Agent Memory Standard
Consent-governed memory pools that let agents share context across systems, so experiences carry over without surrendering control of what's remembered.
01Abstract
Defines how AI agents share memories and context across different systems, enabling continuous and coherent user experiences while maintaining privacy and user control over shared information.
02Key concepts
Typed memories
Memories are classified (preference, context, knowledge, procedure, relationship, temporal), each with confidence, importance, and a lifespan.
Shared pools
Memory pools have explicit read/write access, size limits, and policies for pruning, de-duplication, and merging similar entries.
User control
Owners decide what is shared, with which agents, and for how long; entries can expire by time-to-live or be made permanent deliberately.
Continuity with privacy
Context follows the user across agents through privacy-preserving sync, ending the cycle of re-explaining yourself to every new system.
03Core operations
- createPool(policy) / grantPoolAccess(poolId, agent, mode)
- writeMemory(poolId, memory)
- queryMemory(poolId, filter)
LCS-004 closes the loop: with consent, your AI experiences become continuous across tools. Your assistant in one app can pick up where another left off, on your terms.
04Full specification
The complete, canonical specification for LCS-004 lives in the standards repository and is always the source of truth.
Read on GitHub