The cognition layer for reliable AI agents.
Matrix turns natural language into typed, inspectable Intent IR — so your agents reason deterministically, act safely, and never hallucinate a transaction.
Two rails. One substrate.
Neo — Your everyday co-pilot
The default tool-calling agent for reversible work. Neo handles file edits, research, code generation, and everyday tasks — fast, flexible, and forgiving by design.
MCL — Determinism for high-stakes work
The typed Intent IR for irreversible operations. MCL compiles natural language into inspectable, correctable intent structures before execution — every step is auditable and replayable.
Beyond prompt fragility
Traditional LLMs treat every request as an opaque string. There's no type safety, no inspection, no correction — just a prayer that the model guesses right. One hallucinated parameter and funds vanish or data corrupts irreversibly.