Wavex OS 是 AI Skill Hub 本期精选Agent工作流之一。综合评分 7.5 分，整体质量较高。我们推荐使用将其纳入你的 AI 工具库，帮助提升工作效率。
📚 深度解析
Wavex OS 工作流的设计遵循"最小配置，最大复用"原则：核心逻辑已经封装好，用户只需配置自己的 API Key 和业务参数即可快速上手。工作流内置错误处理和重试机制，在网络波动或 API 限速等情况下仍能稳定运行，适合作为生产环境的自动化基础设施。
在实际部署时，建议先在测试环境中运行 3-5 次，验证各个环节的输出结果符合预期，再部署到生产环境。AI Skill Hub 评分 7.5 分，是同类 Agent 工作流中的精选推荐。
📋 工具概览
Wavex OS 是一套完整的 AI Agent 自动化工作流方案。通过可视化的节点编排，将复杂的多步骤任务拆解为清晰的自动化流程，实现全程无人值守的智能处理。支持与数百种外部服务和 API 无缝集成，适合构建数据处理管线、业务自动化和 AI 辅助决策系统。
📖 中文文档
Wavex OS 是一套完整的 AI Agent 自动化工作流方案。通过可视化的节点编排，将复杂的多步骤任务拆解为清晰的自动化流程，实现全程无人值守的智能处理。支持与数百种外部服务和 API 无缝集成，适合构建数据处理管线、业务自动化和 AI 辅助决策系统。
- 可视化 Agent 工作流编排，无需编写复杂代码
- 支持多步骤自动化任务链，实现全流程无人值守
- 与外部 API、数据库和第三方服务无缝集成
- 内置错误处理与自动重试机制，保障稳定运行
- 提供可复用的自动化模板，快速在同类场景部署
- 自动化日常重复性工作，将精力集中于创造性任务
- 构建数据采集 → 处理 → 输出的完整自动化管线
- 实现跨平台、跨系统的数据流转和业务协同
# 方式一：npm 全局安装 npm install -g wavex-os # 方式二：npx 直接运行（无需安装） npx wavex-os --help # 方式三：项目依赖安装 npm install wavex-os # 方式四：从源码运行 git clone https://github.com/aimerdoux/wavex-os cd wavex-os npm install npm start
- 访问 GitHub 仓库获取工作流文件
- 在对应平台（Dify / Flowise / Make 等）中找到「导入工作流」功能
- 上传工作流文件
- 按照提示配置必要的环境变量和 API Key
- 运行测试确认流程正常后投入使用
# 命令行使用
wavex-os --help
# 基本用法
wavex-os [options] <input>
# Node.js 代码中使用
const wavex_os = require('wavex-os');
const result = await wavex_os.run(options);
console.log(result);
# wavex-os 配置说明 # 查看配置选项 wavex-os --config-example > config.yml # 常见配置项 # output_dir: ./output # log_level: info # workers: 4 # 环境变量（覆盖配置文件） export WAVEX_OS_CONFIG="/path/to/config.yml"
WaveX OS
Open-source operating system for running an AI agent company on your localhost.
WaveX OS is a localhost-first wizard that materializes an AI agent company in about an hour. You answer five questions (the 5 pillars), the wizard picks an agent roster from 165 vetted templates, builds a workflow plan, runs a Monte Carlo simulation against your goal, and activates the fleet into a real Paperclip runtime — where each agent gets heartbeats, KPIs, budgets, and a board.
Your data and inference stay on your machine. Spawned agents use your Claude Max OAuth via a wrapper that reads the system keychain — no API keys, no remote inference, no telemetry.
After the wizard you're on a Mission Control dashboard with a live fleet graph + KPI scoreboard. An optional System Optimizer subscription (Phase F) can inject board-level reasoning on a daily cadence using your same OAuth — code is free, hosted optimizer is a tier on top.
---
What's in the box
wavex-os/
├── apps/installer/ # npx wavex-os init CLI (doctor + spawn dev:full)
├── packages/
│ ├── core/ # Paperclip vendored via git subtree
│ ├── db/ # PGlite (dev) / Postgres (prod) + Drizzle schema
│ ├── plugin-sdk-shim/ # Re-exports @paperclipai/plugin-sdk surface
│ ├── auth-shim/ # assertBoard / assertCompanyAccess gates
│ ├── composio-shim/ # listConnections + featured toolkits
│ ├── inference-adapter/ # tier-router claudeBin (OAuth vs API key mode)
│ ├── wavex-os-server/ # Fastify routes for vendored onboarding + activate + handoff
│ ├── onboarding-ui/ # Vite + React wizard + Mission Control v2
│ │ └── public/agent-templates/ # 165 role-specific skill packs (CEO, CoS, CMO, CRO, …)
│ ├── mock-core/ # In-memory stand-in for Paperclip; Fastify on :3101
│ ├── standard-skills/ # Cross-cutting skills every agent loads
│ ├── healing/ # OAuth refresh + worker restart reference impl
│ ├── observability/ # bottlenecks, attribution, budget, mission-control
│ └── onboarding-server-client/ # Typed stub for future hosted backend
├── vendor/wavex-os/ # vendored wavex-os @ d84983a1 (2026-05-03)
│ ├── plugin-sdk/ # @paperclipai/plugin-sdk
│ ├── shared/ # @paperclipai/shared
│ ├── tier-router/ # @wavex-os/plugin-tier-router
│ ├── flywheel-kernel/ # @wavex-os/plugin-flywheel-kernel
│ ├── onboarding/ # @wavex-os/plugin-onboarding (50 src files)
│ └── VENDOR.md # source SHA + vendor exceptions + update procedure
├── scripts/
│ ├── wrappers/
│ │ ├── claude-anthropic-direct.sh # macOS keychain → OAuth wrapper
│ │ └── claude-spawn.sh # Per-spawn execution wrapper with 401 self-heal
│ ├── render-launchd-templates.mjs # Materialize launchd plists
│ ├── provision-chief-of-staff.sample.mjs # Kernel completion
│ └── setup-hierarchy-and-kpis.sample.mjs # Hierarchy + KPI registration
├── templates/launchd/ # 6 .plist.tmpl with placeholders
├── examples/ # config + KPI registry samples
├── docs/
│ ├── ARCHITECTURE.md
│ ├── CLAUDE_MAX_HANDOFF.md
│ ├── MINIMAL_INCEPTION.md
│ ├── SELF_HEALING.md
│ ├── ROADMAP.md
│ ├── onboarding/migration-plan.md
│ ├── ops/surface-tuning-map.md # 40 tunables (auto-generated)
│ └── images/wizard/ # The screenshots in this README
└── e2e/
├── onboarding.spec.ts # full wizard walk
├── flow-variants.spec.ts # 11 variants (reset, resume, swap, conflict, …)
├── bug-hunt.spec.ts # 14 composition + edge-case scenarios
├── dashboard.spec.ts # Mission Control behavior
├── screenshot-walkthrough.spec.ts # capture for this README
└── screenshot-mission-control.spec.ts---
Quickstart (from source)
You need Node ≥18, pnpm ≥9, git, and a Claude Max subscription on the same machine (the wizard reads from the system keychain).
If you have corepack enabled (corepack enable), pnpm is pinned automatically via the packageManager field — you don't need to install pnpm separately.
git clone https://github.com/aimerdoux/wavex-os.git
cd wavex-os
pnpm install
pnpm -r --filter "./vendor/wavex-os/*" build # build the vendored wavex-os packages
pnpm devpnpm dev also re-runs pnpm install --prefer-offline automatically as a safety net — so if you skip the explicit install step, pnpm dev will still bootstrap cleanly the first time.
Note on a benign warning. Duringpnpm installyou'll see twoWARNlines about apnpmfield inpackages/core/package.json. That's expected —packages/coreis the vendored Paperclip subtree with its own install lifecycle (cd packages/core && pnpm install), and itspatchedDependencies/overridesapply there, not at the outer wavex workspace. Safe to ignore.
This boots two servers in parallel:
http://localhost:5173— onboarding wizard + Mission Control (Vite + React)http://localhost:3101—mock-core(Fastify, hosts/api/*+ wavex-os routes)
Open http://localhost:5173 and you'll land on Mission Control. Click Start onboarding to begin the wizard.
Port collision on 5173
The onboarding UI intentionally uses Vite's strict port mode so the installer, README commands, and Playwright QA flow all point at the same address. If pnpm dev exits because 5173 is already in use, stop the other Vite/dev server and rerun:
```bash lsof -nP -iTCP:5173 -sTCP:LISTEN
Optional — hand off to a running Paperclip instance
If you've got Paperclip running locally, point WaveX OS at it and the wizard's activate step will also hire your C-Suite as real Paperclip agents:
PAPERCLIP_HANDOFF_URL=http://127.0.0.1:3100 \
PAPERCLIP_HANDOFF_WRAPPER=/abs/path/to/claude-anthropic-direct.sh \
pnpm devWithout those env vars, activate writes to the wavex DB only — the agents exist as rows but don't run heartbeats. With them, the full inception loop closes.
npx wavex-os initships fromapps/installer/— Phase F will publish it to npm. For now,pnpm devis the supported path.
Tony Apple QA — quick install
Tony Apple QA is the mobile-app QA distribution of WaveX OS. It gives iOS and Android teams an AI-powered QA agent fleet that runs entirely on your laptop: smoke-test orchestration, regression coverage plans, release checklists, and a live Mission Control dashboard — all backed by your Claude Max subscription, no API keys or cloud accounts required.
npm install -g tony-apple-qa
tony-apple-qa inittony-apple-qa init opens the five-pillar wizard in your browser. Answer questions about your app, your Claude setup, and your QA goals; the wizard builds your agent roster and activates the fleet in under an hour. See packages/tony-apple-qa/ for the full CLI reference.
---
高质量的开源AI工作流操作系统
- 构建多智能体协作系统的 Agent 开发者
- Agent 任务先做 dry-run 验证工具调用链，再开启自主执行
- API key 直接提交到 git 仓库（请用 .env 并加入 .gitignore）
- CLI：直接 npm install -g / pip install，命令行调用
- 云端托管：可放在 Vercel / Railway / Fly.io 等 PaaS 平台
⚡ 核心功能
- 可视化 Agent 工作流编排，无需编写复杂代码
- 支持多步骤自动化任务链，实现全流程无人值守
- 与外部 API、数据库和第三方服务无缝集成
- 内置错误处理与自动重试机制，保障稳定运行
- 提供可复用的自动化模板，快速在同类场景部署
- 构建多智能体协作系统的 Agent 开发者
- Agent 任务先做 dry-run 验证工具调用链，再开启自主执行
- API key 直接提交到 git 仓库（请用 .env 并加入 .gitignore）
👥 适合人群
🎯 使用场景
- 自动化日常重复性工作，将精力集中于创造性任务
- 构建数据采集 → 处理 → 输出的完整自动化管线
- 实现跨平台、跨系统的数据流转和业务协同
⚖️ 优点与不足
- +大幅减少重复性人工操作
- +可视化流程，清晰直观
- +可扩展性强，支持复杂场景
- −初始配置和调试需投入一定时间
- −强依赖外部服务的稳定性
- −复杂场景需具备一定技术基础
该工具使用 NOASSERTION 协议，商用场景请仔细阅读协议条款，必要时咨询法律意见。
AI Skill Hub 为第三方内容聚合平台，本页面信息基于公开数据整理，不对工具功能和质量作任何法律背书。
建议在沙箱或测试环境中充分验证后，再部署至生产环境，并做好必要的安全评估。
📄 NOASSERTION — 请查阅原始协议条款了解具体使用限制。
🔗 相关工具推荐
❓ 常见问题 FAQ
经综合评估，Wavex OS 在Agent工作流赛道中表现稳健，质量良好。如果你已有明确的使用需求，可以直接上手体验；如果还在评估阶段，建议对比同类工具后再做决策。
| 原始名称 | wavex-os |
| 原始描述 | 开源AI工作流：Open-source operating system for running an AI agent company on your own machine。⭐7 · TypeScript |
| Topics | AI工作流操作系统 |
| GitHub | https://github.com/aimerdoux/wavex-os |
| License | NOASSERTION |
| 语言 | TypeScript |
收录时间：2026-05-28 · 更新时间：2026-05-30 · License：NOASSERTION · AI Skill Hub 不对第三方内容的准确性作法律背书。