🧬 OpenEvolant: The Autonomous Evolution Engine

"Stop coding static agents. Start breeding them."

Join us on Discord: https://discord.gg/kueXC3vMKZ

---

Part 1 — English

👤 About Us

Jammie — Former Tencent AiLab, now an AI hardware entrepreneur.
Company: someonewaits.com

Chloe — Co-founder. Ex CVC for Swire Group (太古集团). Strategic and venture perspective. MBA from The University of Sydney.

---

🌟 Overview

OpenEvolant is an open-source autonomous evolution engine built for Skills and MCP. It treats Agent logic not as static code, but as heritable Gens (Genes). By introducing simulated pressure, mutation, and natural selection, OpenEvolant automatically compresses and optimizes redundant reasoning paths into high-performance Apex Skills.

---

⚔️ Comparison

Dimension	Traditional Agents (e.g., static skills)	OpenEvolant
Logic Basis	Static Prompts	Evolving Gens
Optimization	Manual Tuning	Natural Selection
Cost Curve	Consistently High	Collapses via Iteration
Output	Execution results	Evolvable Asset (.gen)

---

🛠️ Ecosystem Role

OpenEvolant acts as a "Parasitic Brain" that enhances existing host frameworks.

• vs. Skills: Skills provide the "Muscles" (task execution). OpenEvolant provides the "Prefrontal Cortex" (logic optimization).
• vs. MCP: MCP provides the "Tools" (static resources). OpenEvolant acts as the "Dispatcher" (learning the most efficient way to call them).

---

🖥️ Evolution GUI

Through Evolant Studio, you can witness digital Darwinism in real-time:

• Mutation Tracking: Watch prompt fragments recombine to find the "optimal sequence."
• Survival of the Fittest: See high-cost, low-accuracy genes being phased out by Natural Selection (.ns) and Evaluator (.evaluator).
• Gene Crystallization: Witness raw logic settling into lightweight, distributable .gen files.

---

📸 Screenshots

Evolution	Genes
Evolution dashboard	Genome browser

---

🚀 Getting Started

1. Clone & Install

git clone <repo-url>
cd openEvolant
npm install

2. Build

npm run build

Builds packages/*, studio, and CLI (@openevolant/cli) in order.

3. Start Web GUI

npm start
# or
node packages/cli/dist/cli.js start
# or (after global install)
openevolant

Starts the local server hosting Evolant Studio; default: http://localhost:3000.
Runtime directory: All config (LLM, .ns natural selection, .evaluator), data (genes, sessions), and skills live under a single app root; default ./openevolant (override with --root=<path>). So both dev and packaged runs use the same layout: openevolant/config/, openevolant/data/, openevolant/skills/.
Options: --port=4000, --root=./openevolant, --host=0.0.0.0.
Help: openevolant --help; version: openevolant --version.

4. Development

npm run dev

Starts the Evolant Studio dev server (Vite) and opens chat / evolution management / settings in the browser.

5. Global Install (TBD)

npm install -g openevolant

---

📚 Documentation

Doc	Description
ARCHITECTURE	Architecture and design
FOLDER_STRUCTURE	Project folder structure
SPEC_genes	Genes (Gens) specification
SPEC_ns	Natural Selection (.ns) specification
SPEC_evaluator	Single-condition Evaluator (.evaluator) specification
VERSIONS	Version notes

---

Part 2 — 中文

👤 关于我们

Jammie — 前腾讯 AiLab，目前 AI 硬件创业者。
公司： someonewaits.com

---

🌟 项目概览

OpenEvolant 是专为 Skills 与 MCP 打造的开源自主进化引擎。它不将 Agent 逻辑视为死板的代码，而是视其为可遗传的 Gens（基因）。通过引入模拟压力、变异与自然选择，OpenEvolant 能够自动将冗余的推理路径压缩并优化为高性能的 Apex Skills（巅峰技能）。

---

⚔️ 深度对比

维度	传统 Agent（如静态 Skills）	OpenEvolant
逻辑本质	静态指令	进化基因
优化方式	手动调优	自然选择
成本曲线	持续高昂	随迭代坍缩
核心产物	执行结果	进化资产（.gen）

---

🛠️ 生态位定位

OpenEvolant 作为一个「寄生大脑」，负责增强现有的宿主框架。

• 与 Skills：Skills 提供「肌肉」（任务执行），OpenEvolant 提供「前额叶」（逻辑优化）。
• 与 MCP：MCP 提供「工具」（静态资源），OpenEvolant 充当「调度员」（学习最高效的调用方式）。

---

🖥️ 进化可视化

通过 Evolant Studio，你可以实时见证数字达尔文主义：

• 变异追踪：观察 Prompt 碎片如何重组以寻找「最优序列」。
• 优胜劣汰：观察高成本、低准确率的基因如何被自然选择（.ns）与评估器（.evaluator）淘汰。
• 基因沉淀：见证原始逻辑如何凝结为轻量、可分发的 .gen 文件。

---

📸 界面预览

进化	基因组
进化看板	基因组浏览

---

🚀 快速开始

1. 克隆与安装

git clone <repo-url>
cd openEvolant
npm install

2. 构建

npm run build

会依次构建 packages/*、studio 和 CLI（@openevolant/cli）。

3. 一键启动 Web GUI

npm start
# 或
node packages/cli/dist/cli.js start
# 或（全局安装后）
openevolant

启动本地服务，托管 Evolant Studio 静态资源；默认 http://localhost:3000。
运行时目录：配置（LLM、.ns 自然选择、.evaluator）、数据（genes、会话）、skill 等统一放在一个应用根目录下，默认 ./openevolant（可用 --root=<路径> 覆盖）。开发与打包后运行布局一致：openevolant/config/、openevolant/data/、openevolant/skills/。
选项：--port=4000、--root=./openevolant、--host=0.0.0.0。
帮助：openevolant --help；版本：openevolant --version。

4. 开发

npm run dev

启动 Evolant Studio 开发服务器（Vite），在浏览器中打开聊天 / 进化管理 / 设置占位页。

5. 全局安装（待发布）

npm install -g openevolant

---

📚 Docs 快速跳转

文档	说明
ARCHITECTURE	架构与设计
FOLDER_STRUCTURE	项目目录结构
SPEC_genes	基因（Gens）规范
SPEC_ns	自然选择（.ns）规范
SPEC_evaluator	单条件评估器（.evaluator）规范
VERSIONS	版本说明