Tools for coherent LLM coding agents.
Persistent memory and multi-agent orchestration, modeled on how biological minds actually work.
cerebral.work · reverie · meshctl · docs
A memory daemon modeled on human consolidation. Observation-layer writes, typed-then-distilled reads, sleep-phase consolidation between sessions. Ports 70 years of memory-consolidation research into the daily workflow of building software with AI.
Multi-agent orchestration with file-level locking and coord-protocol messaging. Runs many Claude Code / Cursor / Windsurf sessions in parallel without races. Replaces ad-hoc tmux hacks with a distributed-systems substrate.
LLM coding harnesses accumulate knowledge across sessions but lack a theory of where it should live.
We’re fixing that.