by ljxpython
企业级 AI Agent 平台底座,基于 LangGraph 生态体系二次开发,开箱即用(FastAPI + Vue 3 + MCP + Skills + 沙箱工作区 + 长期记忆)
# Add to your Claude Code skills
git clone https://github.com/ljxpython/ai-agent-platformGuides for using ai agents skills like ai-agent-platform.
Last scanned: 9/30/2026
{
"issues": [],
"status": "PASSED",
"scannedAt": "2026-09-30T10:18:44.925Z",
"npmAuditRan": true,
"pipAuditRan": true,
"promptInjectionRan": true
}See how ai-agent-platform compares with popular alternatives.
ai-agent-platform is an open-source ai agents skill for AI coding assistants such as Claude Code, Codex CLI, and ChatGPT, built by ljxpython. 企业级 AI Agent 平台底座,基于 LangGraph 生态体系二次开发,开箱即用(FastAPI + Vue 3 + MCP + Skills + 沙箱工作区 + 长期记忆). It has 134 GitHub stars.
Yes. ai-agent-platform passed SkillsLLM's automated security scan — a dependency vulnerability audit plus prompt-injection heuristics — with no high-severity issues. You can read the full report in the Security Report section on this page.
Clone the repository with "git clone https://github.com/ljxpython/ai-agent-platform" and add it to your Claude Code skills directory (see the Installation section above).
ai-agent-platform is primarily written in Python. It is open-source under ljxpython on GitHub, so you can review or fork the full source.
Yes. SkillsLLM lists many other AI Agents skills you can browse and compare side by side. Open the AI Agents category from the badge at the top of this page, or use the Related Skills and comparison links further down to weigh ai-agent-platform against similar tools.
No comments yet. Be the first to share your thoughts!
⚠️ Third-Party Software Notice
This skill is third-party open-source software developed and hosted independently on GitHub. SkillsLLM is an informational directory and does not control or maintain the underlying repository.
Any security checks, ratings, or warnings displayed by SkillsLLM are automated and limited in scope. They do not constitute a security certification or guarantee that the software is safe, error-free, or free from malicious code, vulnerabilities, compromised dependencies, or prompt-injection risks.
Review the source code, permissions, dependencies, and configuration before installing or running any third-party skill. Use is at your own risk. To the maximum extent permitted by applicable law, SkillsLLM is not liable for losses arising from third-party software.
See comparison
很多团队做 Agent 容易停留在 Demo 阶段:平台治理、运行时执行、状态存储和前端交互全揉成一个“大泥球”,一到生产落地就面临权限缺失、状态丢失、模型切换困难、二开举步维艰的问题。
本项目为解决这一痛点而生,提供一个可直接用于二次开发、快速搭建企业私有 AI Agent 平台的工程底座:
LangGraph / LangChain 系列生态设计,深度吸收 open-swe、deepagents 与 deer-flow 的工程思想,原生支持复杂图编排、状态持久化、多轮工具调用循环与人机协同(HITL)审批中断。整个系统由三大服务分层构建,各司其职:
| 服务 | 目录 | 职责定位 | 核心技术栈 |
|---|---|---|---|
| Platform API | apps/platform-api |
控制面核心:认证鉴权、项目治理、审计日志、模型 Catalog 目录、受管契约网关转发 | FastAPI + SQLAlchemy + PostgreSQL |
| Platform Web | apps/platform-web |
管理控制台前端:工作台布局、Agent 交互对话流、权限管理、多端流式渲染 | Vue 3 + Vite + Tailwind CSS + Pinia |
| Runtime Service | apps/runtime-service |
Agent 执行引擎:LangGraph 图注册、工具与 MCP 装配、会话调度、SSE 事件流保活推送 | Python 3.11+ + LangGraph + GraphHarbor + Redis |
🔗 交互式网页体验: 👉 打开全屏交互式架构图 (HTML)(支持节点聚焦缩放、深浅主题切换与全要素搜索)

🔗 交互式网页体验: 👉 打开全屏交互式时序图 (HTML)(支持三阶段分段探索与完整链路追踪)

🔗 交互式网页体验: 👉 打开全屏交互式扩展点图 (HTML)(清晰划定二开边界)

仓库内置了两套具有不同使命的智能体示例,兼顾了快速上手学习与极端复杂场景落地:
showcase_demoapps/runtime-service/src/runtime_service/services/demo/showcase_demo/DeerFlow Agentbytedance/deer-flow、langchain-ai/deepagents 与 langchain-ai/open-swe 的工程设计范式。💡 企业二开建议: 企业客户可以直接复用
DeerFlow Agent的工程实现作为高阶智能体模板,也可将其拆解为底层组件,按需装配到企业原有的垂直业务场景中。
| 扩展需求 | 目标代码路径 | 开发说明 |
|---|---|---|
| 新增自定义 Agent 图 | apps/runtime-service/src/runtime_service/graphs/ |
基于 LangGraph 编写 StateGraph,定义节点与边,并在统一入口注册 |
| 新增自定义工具 (Tools) | apps/runtime-service/src/runtime_service/tools/ |
使用 @tool 装饰器编写纯 Python 函数,平台自动提取 JSON Schema 供模型调用 |
| 接入第三方 MCP 服务 | apps/runtime-service 配置文件 |
标准 MCP 客户端开箱即用,通过配置快速挂载外部 FastMCP / 官方 MCP 工具服务 |
| 扩展控制面 API | apps/platform-api/src/platform_api/ |
遵循 apps/platform-api/docs/handbook/ 规范新增 REST 端点与数据模型 |
| 管理台页面二次开发 | apps/platform-web/src/modules/ |
遵循 control-plane-page-standard.md 页面标准,快速扩建控制台视图 |
当前平台控制台已完成多次大版本重构,消灭了早期简陋的 Demo 样貌,全面演进为现代化工业级控制台:

平台当前已沉淀出 4 大核心视觉与交互空间:
💡 关于界面体验补充:我们正在准备一套完整的 30 秒快速漫游短视频与高帧率 GIF 动图,欢迎保持关注!
# 1. 激活 runtime 虚拟环境(确保依赖已安装)
source "apps/runtime-service/.venv/bin/activate"
# 2. 运行健康自检,检查数据库、Redis 连接与配置文件
bash "scripts/local-stack.sh" doctor
# 3. 自动执行数据库迁移并拉起全栈进程 (Runtime API、Worker、Platform API、Platform Web)
bash "scripts/local-stack.sh" start
# 4. 查看当前栈运行状态与端口占用
bash "scripts/local-stack.sh" status
# 5. 停止本地全栈服务
bash "scripts/local-stack.sh" stop
首次运行的配置初始化、数据库建表及密码配置,请查阅 非容器化本地部署手册。
# 选项 A:仅启动 runtime-service 执行层
docker compose -f apps/runtime-service/deploy/docker-compose.runtime-service.yml \
--env-file apps/runtime-service/deploy/.env.runtime-service up -d
# 选项 B:启动全栈 stack(前后端独立暴露端口)
docker compose -f deploy/docker-compose.stack.yml --env-file deploy/.env.stack up -d
# 选项 C:启动全栈 stack(带 Nginx 反向代理,单端口统一入口)
docker compose -f deploy/docker-compose.stack.nginx.yml --env-file deploy/.env.stack up -d
完整容器指南见 deploy/README.md 与 容器化零到一运行指南。
| 模块 | 默认本地地址 | 最小健康检查指令 |
|---|---|---|
| Platform Web | http://127.0.0.1:3000 |
浏览器直接访问前端管理界面 |
| Platform API | http://127.0.0.1:2142 |
curl -fsS "http://127.0.0.1:2142/_system/health" |
| Runtime Service | http://127.0.0.1:8123 |
curl -fsS "http://127.0.0.1:8123/ready" |
ai-agent-platform/
├── apps/
│ ├── platform-api/ # 平台控制面后端 (FastAPI, 权限/项目/审计/Catalog)
│ ├── platform-web/ # 平台控制面前端 (Vue 3, 统一管理台与聊天流)
│ └── runtime-service/ # LangGraph 执行运行时 (图编排/工具装配/状态机)
├── deploy/ # Docker Compose 生产与开发镜像编排
├── docs/ # 架构设计、场景指南与跨服务标准体系
│ ├── architecture/ # 系统架构沉淀与概念透析专篇
│ ├── diagrams/ # 交互式架构与时序图表 (Archify HTML)
│ ├── guides/ # 开发者指南、部署手册与数据库运维规范
│ └── standards/ # 跨服务通信、错误信封与追踪标准
├── scripts/ # 本地栈启停管理与一致性检查脚本
└── AGENTS.md # 团队工程规范与 AI 协同开发指南
v0.5.0(迭代记录见 CHANGELOG.md)如果你在企业内部落地 Agent 平台、使用 LangGraph 进行二次开发或使用 MCP 扩展能力时遇到问题,欢迎交流探讨:
个人微信号:
本项目在持续演进过程中,深度受益于以下开源项目与核心工程思想: