Cognitive agent company
Percena builds the full stack behind cognitive agents: real-time perception, durable memory, and the governance that lets them work alongside people and earn trust inside your products and teams.
The Percena stack
Percena stands for Perception, Cognition, and Persona. Q-Vis handles real-time perception, MemFuse provides durable agent memory, Weft embeds agents into software products, and Codebreeze orchestrates multiple agents across engineering work.
The stack is already connected in code: Codebreeze uses MemFuse memory and the Weft SDK.
Four scenes
观 · Perceive
Voice-first agents that listen, reason, and respond in real time.
Two-mode dialogue · Fast/slow dual-track · Semantic barge-in
忆 · Remember
Persistent agent memory that retains what matters and forgets the rest.
Ebbinghaus forgetting · Signal-beacon retrieval · Deterministic fallback
织 · Weave
Embed AI agents into your product without rebuilding your backend.
OpenAPI → tools · Replayable timeline · Multi-tenant metering
合 · Coordinate
Run multiple coding agents like one coordinated team — on a shared, auditable workflow.
Ticket-driven · Overlay conflict detection · Git consensus
Why Percena
An agent begins by knowing the people it serves, then turns inward to know itself and renew itself, until it stands on its own. We, in turn, meet it as a partner, and work alongside it. Two paths, drawn toward each other.
Perception, memory, runtime, and orchestration are designed to interoperate from the start, not assembled from separate tools.
MemFuse and the Weft SDK are MIT-licensed, so teams can adopt the core layers first and move to managed control planes when ready.
Approval gates, replayable timelines, audit chains, and conflict detection keep agent work visible and traceable — so you can trust it with real work.
Tell us what your product or team needs agents to see, remember, or do — we will map it to the right layer of the Percena stack.
