经 AI Skill Hub 精选评估,多智能体框架 获评「强烈推荐」。这款Agent工作流在功能完整性、社区活跃度和易用性方面表现出色,AI 评分 8.0 分,适合有一定技术背景的用户使用。
多智能体框架 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。
多智能体框架 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 kiro-multi-agents
# 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 kiro-multi-agents
# 方式三:从源码安装(获取最新功能)
git clone https://github.com/vankhangfet/kiro-multi-agents
cd kiro-multi-agents
pip install -e .
# 验证安装
python -c "import kiro_multi_agents; print('安装成功')"
# 命令行使用
kiro-multi-agents --help
# 基本用法
kiro-multi-agents input_file -o output_file
# Python 代码中调用
import kiro_multi_agents
# 示例
result = kiro_multi_agents.process("input")
print(result)
# kiro-multi-agents 配置文件示例(config.yml) app: name: "kiro-multi-agents" debug: false log_level: "INFO" # 运行时指定配置文件 kiro-multi-agents --config config.yml # 或通过环境变量配置 export KIRO_MULTI_AGENTS_API_KEY="your-key" export KIRO_MULTI_AGENTS_OUTPUT_DIR="./output"
A production-ready multi-agent configuration for Kiro CLI that automates the full software development lifecycle — from business problem discovery to deployed, reviewed, and documented code — using a team of specialized AI agents that collaborate autonomously under human oversight.
---
User: "I want to build an internal expense approval system"
│
└─► PM: 8-category discovery questions (business problem, users, features,
│ flows, NFRs, integrations, UI, launch criteria)
│
└─► User: answers
│
└─► PM: Option A — "Lightweight form + email approval"
│ Option B — "Full workflow platform with audit trail"
│
└─► User: "Option B, but start with 2 approval levels only"
│
└─► PM: Writes PRD, shows requirements summary
│
└─► User: YES
│
└─► Architect: spec.md + tasks.md
├─► UI/UX: screen mockups
├─► Coder: implementation groups (sequential)
├─► Reviewer: PASS/FAIL gate
├─► Security Reviewer: PASS/FAIL gate
└─► Docs: arc42 + C4 + README update
The PM runs first, always. No agent creates spec.md or tasks.md until the user has confirmed the PRD. This prevents the most common failure mode in AI-assisted development: building the wrong thing confidently and completely.
```bash
After gathering requirements, the PM proposes two meaningfully different approaches — not just "simple vs complex", but trade-offs that reflect the user's real constraints (team size, timeline, budget, expected scale). The user chooses the direction before any architecture work begins.
Triggered by Architect immediately after planning, before any implementation.
npx uipro-cli init --ai kiroui/screens/NN-name.html per screen, ui/transitions/flow.html (animated flow), ui/index.html (navigation hub), ui/design-system.mdModel: claude-sonnet-4.6 | MCPs: context7
---
cd /path/to/your-project kiro ```
The pm agent is the default agent and starts automatically.
┌─────────────────────────────────────────────────────────────────────┐
│ USER REQUEST │
└───────────────────────────────┬─────────────────────────────────────┘
│
▼
┌─────────────────────────────────┐
│ PM AGENT │
│ │
│ Turn 1: Discovery questions │◄── User answers
│ Turn 2: 2 solution options │◄── User chooses
│ Turn 3: PRD + summary │◄── User confirms
│ Turn 4: Dispatch │
└─────────────────┬───────────────┘
│ confirmed PRD
▼
┌─────────────────────────────────┐
│ ARCHITECT AGENT │
│ │
│ Reads PRD → writes spec.md │
│ Writes tasks.md (groups) │
│ Runs execution loop │
└──┬──────────────────────────────┘
│
┌─────────┼─────────────────────────────────┐
│ │ │
▼ ▼ ▼
┌─────────┐ ┌─────────┐ ┌───────────┐
│ UI/UX │ │ CODER │ ··· (per group) │ OPS │
│ mockups │ │ impl. │ │ infra │
└─────────┘ └─────────┘ └───────────┘
│ all groups complete
▼
┌──────────────────┐ ┌──────────────────┐
│ REVIEWER │────►│ SECURITY REVIEWER │
│ code quality │ │ OWASP / secrets │
└──────────────────┘ └──────────────────┘
│ PASS
▼
┌──────────────────┐
│ DOCS AGENT │
│ arc42 + C4 + README│
└──────────────────┘
---
The PM asks across eight categories, phrased for business stakeholders — not developers:
| # | Category | What it uncovers |
|---|---|---|
| 1 | **Business Problem** | Pain, cost of inaction, success definition, v1 boundaries |
| 2 | **Users & Personas** | Who uses it, how many, what they do today without it |
| 3 | **Functional Core** | Actions users take, permissions, data model, business rules |
| 4 | **Flows & Edge Cases** | Main journey, secondary paths, error states, empty states |
| 5 | **Quality & Reliability** | Performance targets, uptime SLA, auth method, compliance |
| 6 | **Integrations** | Existing systems, third-party APIs, stack, hard constraints |
| 7 | **UI & Screens** | Screen list, platform, design system, UX patterns |
| 8 | **Launch & Sign-off** | Acceptance criteria, test data, approver, rollout plan |
Sample questions in business language: - "What's the cost of NOT having this? Time lost, mistakes made, revenue not captured?" - "How do people cope today — spreadsheet, manual process, another tool, or nothing?" - "How fast does it need to feel? Is a 2-second load acceptable, or does it need to be instant?" - "Who signs off before this ships?"
---
高质量的自动化工作流框架
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。
AI Skill Hub 点评:多智能体框架 的核心功能完整,质量优秀。对于自动化工程师和运维人员来说,这是一个值得纳入个人工具库的选择。建议先在非生产环境试用,再逐步推广。
| 原始名称 | kiro-multi-agents |
| 原始描述 | 开源AI工作流:Production-ready multi-agent framework for Kiro CLI that automates the end-to-en。⭐4 · Python |
| Topics | agent-orchestrationcoding-agentskiro-cli-multi-agent |
| GitHub | https://github.com/vankhangfet/kiro-multi-agents |
| 语言 | Python |
收录时间:2026-06-22 · 更新时间:2026-06-22 · 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