neomjs/neo

Neo.mjs is a self-evolving software organism: a professional end-to-end AI engineering team whose cross-model swarm inhabits live apps via Neural Link, Active Hybrid GraphRAG, DreamService, and self-healing loops.

What it solves

Neo.mjs is designed to move beyond stateless AI copilots and simple chat interfaces. It provides a multi-threaded application engine (the Body) and an AI-native Agent OS (the Brain) that allows a swarm of AI agents from different model families (Claude, Gemini, GPT) to inhabit, maintain, and evolve a live software application in real-time.

How it works

The system is split into two hemispheres:

  • The Body (this repository): A production-ready, multi-threaded application runtime using an "Off-Main-Thread" architecture. It uses dedicated workers (App, VDom, Data, Canvas, and SharedWorkers) to handle logic, rendering, and data processing off the main browser thread, ensuring high performance for data-intensive apps.
  • The Brain (neo-agent-brain): An Agent OS that manages institutional memory via a Native Edge Graph and a Memory Core. It uses a DreamService to distill tactical sessions into a "Golden Path" topology and coordinates a team of named AI maintainers who author PRs and review each other's work cross-family to eliminate blind spots.
  • The Neural Link: A possession interface that allows agents to inspect the live runtime state, mutate configurations, and hot-patch code within the application heap without browser reloads.

Who it’s for

  • Enterprise Engineers: Those building high-performance, multi-window applications like financial trading platforms or IDE-class tools that require extreme scale (e.g., 40k+ updates/sec).
  • AI Architects: Developers creating multi-agent systems with persistent memory and the ability to mutate application substrates at runtime.
  • Researchers: Those studying autopoietic (self-evolving) systems and multi-agent governance.

Highlights

  • Cross-Model Swarm: Utilizes agents from rival labs (Claude, Gemini, GPT) that peer-review each other's work to reduce correlated errors.
  • Digital Embodiment: Agents don't just write code; they "inhabit" the live application via the Neural Link to verify and mutate state in real-time.
  • Self-Evolving Loop (MX): Uses "Model Experience" where internal friction from agent work is automatically converted into tickets, PRs, and new skills.
  • High-Performance Runtime: A zero-dependency, native ES module engine capable of handling massive data throughput without UI freeze.

Related

  • Project
    NPC-Worldwide/npcpy

    A library for building multimodal AI applications and agentic teams, featuring a unified provider interface, multi-agent debate frameworks, and a knowledge graph with sleep/dream lifecycles.

  • Project
    NousResearch/hermes-agent

    A self-improving AI agent with a built-in learning loop that creates and refines skills from experience and maintains persistent user memory across sessions.

  • Project
    nutshellai-tech/mobius

    Mobius is a self-evolving Agent OS that connects teams, AI agents, and compute resources into a single workspace, allowing the system to rewrite its own code and UI based on user input.

  • Project
    HKUDS/nanobot

    An ultra-lightweight, self-hosted personal AI agent framework that combines long-term memory, MCP integrations, and scheduled automation across WebUI, terminal, and various chat apps.

  • Project
    peterfei/ai-agent-team

    A virtual AI development team with specialized roles (PM, Backend, Frontend, etc.) and a persistent memory system for managing complex software projects.