Last released Apr 8, 2026
An SDK for sharing KV-cache and attention states across LLM agents, enabling zero-token handoffs with cross-model projection support.
Supported by