VaultysClaw 是 AI Skill Hub 本期精选Agent工作流之一。综合评分 8.0 分,整体质量较高。我们强烈推荐将其纳入你的 AI 工具库,帮助提升工作效率。
VaultysClaw 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。
VaultysClaw 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。
# 方式一:npm 全局安装 npm install -g vaultysclaw # 方式二:npx 直接运行(无需安装) npx vaultysclaw --help # 方式三:项目依赖安装 npm install vaultysclaw # 方式四:从源码运行 git clone https://github.com/vaultys/VaultysClaw cd VaultysClaw npm install npm start
# 命令行使用
vaultysclaw --help
# 基本用法
vaultysclaw [options] <input>
# Node.js 代码中使用
const vaultysclaw = require('vaultysclaw');
const result = await vaultysclaw.run(options);
console.log(result);
# vaultysclaw 配置说明 # 查看配置选项 vaultysclaw --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 VAULTYSCLAW_CONFIG="/path/to/config.yml"
🔐 Zero-trust foundation - VaultysId — each agent & user gets a unique, cryptographically-rooted identity (a passport, not a password) - Signed intents — all agent work is cryptographically signed; no one can deny what they did - Policy engine — express rules without code: allow read_database if time >= 9am AND time <= 5pm AND day != weekend - Capabilities, not credentials — agents are granted capabilities, never raw API keys
🎛️ Governance out of the box - Workspaces — multi-tenant namespaces for teams, departments, or customers - Token budgets — daily/monthly spend limits per agent and workspace, tracked in real time - Approval workflows — visual editor routes high-risk intents to humans first - Governance dashboard — live posture: agent coverage, policy violations, budget spend, risk score - Audit trail — every operation (who, what, when, why) with cryptographic proof
🚀 Agent orchestration - Workflow engine — drag-and-drop visual editor (React Flow) with sequential/parallel execution, loops, and human approval steps - 30+ skills & tools — file ops, shell, HTTP, code runner, remote agents, knowledge search, Slack, email - Semantic memory — persistent agent memory with auto-summarization and retrieval - Multi-agent coordination — agents delegate via cryptographically-signed peer grants - LiteLLM registry — centrally manage any LLM (OpenAI, Anthropic, Llama, local); agents pick from an approved roster
🔌 Enterprise integrations - Microsoft Entra ID — auto-sync users and groups from Azure AD (groups map to workspaces) - SMTP — configurable email notifications - Docker Compose — pre-built dev + production stacks - LiteLLM proxy — route requests to any model provider
---
git clone https://github.com/vaultys/VaultysClaw.git
cd VaultysClaw
./quick-start.sh # checks docker/node/npm/openssl, generates .env, builds & starts the stack
Then open http://localhost:3000/quick-start — a guided page that logs you in (developer mode, no VaultysID app needed), so you can create an agent, assign capabilities, and watch every action get audited.
⚠️ Quick-start login (VC_DEV_LOGIN=true) is for local testing only — never use it in production.
<details> <summary><b>Manual setup (pnpm)</b></summary>
Requirements: Node.js 18+, pnpm 10+
git clone https://github.com/vaultys/VaultysClaw.git
cd VaultysClaw
pnpm install
pnpm dev
This boots everything: - ✓ Control plane UI → http://localhost:3000 - ✓ Agent web dashboard → http://localhost:3002 - ✓ WebSocket server → ws://localhost:8080 - ✓ Passwordless QR-code login (scan to create your first identity)
</details>
---
---
One command boots the whole stack — control plane, agents, and real-time audit.
---
| Capability | VaultysClaw | API keys | OAuth | Traditional RBAC |
|---|---|---|---|---|
| Non-transferable identity | ✅ | ❌ | ⚠️ | ❌ |
| Deny-by-default permissions | ✅ | ❌ | ❌ | ⚠️ |
| Cryptographic proof of who-did-what | ✅ | ❌ | ✅ | ❌ |
| Policy-driven (not code-driven) | ✅ | ❌ | ❌ | ⚠️ |
| Sub-agent isolation | ✅ | ❌ | ❌ | ⚠️ |
| Real-time approval workflows | ✅ | ❌ | ❌ | ❌ |
| Open-source, self-hosted | ✅ | ⚠️ | ⚠️ | ✅ |
---
高质量的AI工作流编排项目
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。
✅ MIT 协议 — 最宽松的开源协议之一,可自由商用、修改、分发,仅需保留版权声明。
经综合评估,VaultysClaw 在Agent工作流赛道中表现稳健,质量优秀。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 | VaultysClaw |
| Topics | aiai-agentsai-governanceself-hosted |
| GitHub | https://github.com/vaultys/VaultysClaw |
| License | MIT |
| 语言 | TypeScript |
收录时间:2026-07-09 · 更新时间:2026-07-09 · License:MIT · AI Skill Hub 不对第三方内容的准确性作法律背书。
选择 Agent 类型,复制安装指令后粘贴到对应客户端