多AI代理云原生 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。
多AI代理云原生 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。
多AI代理云原生 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:pip 安装(推荐)
pip install multi-ai-agents-cloud-native
# 方式二:虚拟环境安装(推荐生产环境)
python -m venv .venv
source .venv/bin/activate # Windows: .venv\Scripts\activate
pip install multi-ai-agents-cloud-native
# 方式三:从源码安装(获取最新功能)
git clone https://github.com/kinfey/Multi-AI-Agents-Cloud-Native
cd Multi-AI-Agents-Cloud-Native
pip install -e .
# 验证安装
python -c "import multi_ai_agents_cloud_native; print('安装成功')"
# 命令行使用
multi-ai-agents-cloud-native --help
# 基本用法
multi-ai-agents-cloud-native input_file -o output_file
# Python 代码中调用
import multi_ai_agents_cloud_native
# 示例
result = multi_ai_agents_cloud_native.process("input")
print(result)
# multi-ai-agents-cloud-native 配置文件示例(config.yml) app: name: "multi-ai-agents-cloud-native" debug: false log_level: "INFO" # 运行时指定配置文件 multi-ai-agents-cloud-native --config config.yml # 或通过环境变量配置 export MULTI_AI_AGENTS_CLOUD_NATIVE_API_KEY="your-key" export MULTI_AI_AGENTS_CLOUD_NATIVE_OUTPUT_DIR="./output"

A collection of multi-agent AI application samples designed for cloud-native deployment on Microsoft Azure. This repository demonstrates how to build, orchestrate, and deploy intelligent AI agent systems using modern cloud technologies.
Multi-agent systems represent the next evolution in AI applications, where specialized agents collaborate to solve complex tasks. This repository provides practical examples of building such systems with:
Before running any example, ensure you have:
make up
chmod +x setup.sh && ./setup.sh docker compose run --rm podcast-app
#### Monitoring
bash
bash infra/02-build-push.sh
bash infra/03-deploy.sh
bash infra/02-install-airunway.sh
bash infra/03-deploy-qwen.sh kubectl -n airunway-models wait --for=condition=Ready modeldeployment/llama3-2-1b-cpu --timeout=20m
bash infra/05-deploy-agents.sh
---
cp .env.example .env vim .env # fill in SERPAPI_KEY
📁 Location: code/AKS_MicroVM/
A hardened GitHub Copilot SDK Agent service running on Azure Kubernetes Service (AKS) with Kata Containers microVM isolation (kata-vm-isolation). Each pod runs inside an isolated Microsoft Hyper-V (mshv) lightweight VM with its own guest kernel, drastically reducing the blast radius of container escape when the Agent executes untrusted, model-generated code (shell, file I/O, MCP servers, npx packages).
| Layer | Protection |
|---|---|
| **Pod sandbox** | runtimeClassName: kata-vm-isolation → microVM + isolated guest kernel |
| **Container** | runAsNonRoot, readOnlyRootFilesystem, drop ALL caps, seccompProfile: RuntimeDefault |
| **Network** | NetworkPolicy restricts egress to required Copilot / GitHub / MCP endpoints |
| **Secrets** | GH_TOKEN via Kubernetes Secret (swappable with CSI + Azure Key Vault) |
| **Agent tools** | on_permission_request deny-by-default with explicit allowlist |
| Feature | Description |
|---|---|
| **Kata microVM Isolation** | Each pod runs in its own Hyper-V lightweight VM with a dedicated guest kernel |
| **Microsoft Agent Framework + Copilot SDK** | FastAPI service wrapping GitHubCopilotAgent with sync and streaming endpoints |
| **Untrusted Code Containment** | Safe to run Copilot CLI, MCP servers, and arbitrary npx packages |
| **NetworkPolicy Egress Control** | Only required outbound destinations allowed |
| **Defense-in-Depth Pod Security** | Non-root, read-only root FS, dropped caps, seccomp RuntimeDefault |
| **AKS Pod Sandboxing** | Uses AKS-native kata-vm-isolation RuntimeClass on Azure Linux nodes |
Standard_D4s_v3)```bash cd code/AKS_MicroVM
kubectl proxy --port=8001 & curl -s -X POST \ http://localhost:8001/api/v1/namespaces/copilot-agent/services/copilot-agent:80/proxy/chat \ -H 'content-type: application/json' \ -d '{"message":"Briefly introduce Kata Containers."}' ```
⚠️ Kata caveat: kubectl port-forward does not work against Kata pods because the listener lives inside the microVM, not in the sandbox netns. Use the API server proxy, an in-cluster client, or expose the Service via Ingress / LoadBalancer.
---
bash infra/06-show-mcp-endpoints.sh ```
After deployment, update the bundled code/BYOT_Dev/.vscode/mcp.json with the LoadBalancer IPs printed by step 6. In Copilot Chat agent mode, you can ask: "Use the byot tower to take this idea - a URL shortener with click analytics - from requirements through deployment."
---
高质量的AI工作流开源项目
该工具未明确声明开源协议,商业使用前请联系原作者确认授权范围,避免侵权风险。
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。
经综合评估,多AI代理云原生 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 | Multi-AI-Agents-Cloud-Native |
| 原始描述 | 开源AI工作流:These are multi-ai-agents samples for Cloud Native。⭐21 · Python |
| Topics | AICloud NativePython |
| GitHub | https://github.com/kinfey/Multi-AI-Agents-Cloud-Native |
| 语言 | Python |
收录时间:2026-06-01 · 更新时间:2026-06-01 · License:未公布 · 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