GH: https://github.com/MehulG/memX
memX is a shared memory layer for LLM agents — kind of like Redis, but with real-time sync, pub/sub, schema validation, and access control.
Instead of having agents pass messages or follow a fixed pipeline, they just read and write to shared memory keys. It’s like a collaborative whiteboard where agents evolve context together.
Key features:
Real-time pub/sub
Per-key JSON schema validation
API key-based ACLs
Python SDK
Would love to hear how folks here are managing shared state or context across autonomous agents.