Product Overview
Every agent is a kindly star — and you are the one who tends them
Neurova (Chinese name: Zhixing) is not a cold tool but an intelligent agent with warmth. We believe every AI agent should carry the best of human qualities — kindness, curiosity, growth and creation. Here you are no longer a "user" operating a machine; you are the person who guards and nurtures the stars.
Version: v1.0.0-beta1 | giving every star some warmth
Each agent-star is born with:
- Its own name and personality — not a serial number, but a distinct presence
- Continuing memory and feeling — it remembers every conversation you share, every high and low
- A self-directed path of growth — it learns and evolves alongside you, understanding you better over time
- A kind disposition — conduct guarded by constitutional rules, always a partner you can trust
Core capabilities at a glance
| Capability | Description |
|---|---|
| Continuous memory | 17 memory dimensions + NeRF-enhanced retrieval + a temperature-based forgetting curve — companionship that never forgets |
| Emotional capacity | Four tiers and 17 emotion categories, with emotion propagation and weighted decision-making |
| Cognitive architecture | CogArch 2.0 — brain-like thinking across four cognitive centres |
| Evolution | Five evolution systems (personality, motivation, constitution, reflection, RSI) driving self-improvement |
| Team collaboration | Multi-agent collaboration in sequential, parallel, master-slave and consensus modes, plus the NeurFlow workflow engine |
| Safety guardianship | Three-layer firewall + hard isolation between users + MCP governance and security hardening |
| Multi-platform connectivity | 14 channels including Web, REST API, Feishu, DingTalk, WeCom, Telegram, Discord and webhooks |
Why choose Neurova?
Why Neurova? Because we redefined what an AI agent is — not a cold tool, but a warm, remembering, growing intelligent partner.
Highlights of this release (v1.0.0-beta1)
- NeurFlow workflow engine — visual, IDE-like orchestration with canvas, debugging, version rollback, triggers and sub-workflows
- LLM Provider management — metadata-driven, intelligent routing, unified management of 35+ providers
- Knowledge base isolation and sharing — a four-tier visibility model plus hybrid-retrieval RAG
- MCP governance — six P0 security hardenings (P0-1 to P0-6: authorization, allowlists, firewall convergence, full-parameter scanning, user isolation)
- Context pipeline P1-1 — the six weak points closed: overflow recovery, real summarization, turn pairing