IntentOrch
意图驱动的 MCP 编排引擎 — 用自然语言，驱动一切
中文 | English
______________________________________________________________________
🌟 a one-sentence introduction
IntentOrch It belongs to the MCP (Model Context Protocol) ecosystemIntent driven orchestration engineYou only need natural language to describe the requirements, and it can automatically understand intent, match tools, choreograph workflows, and execute them.
💡 Core conceptFrom 'telling the computer how to do it' to 'telling the computer what you want'
______________________________________________________________________
🎯 Why do we need IntentOrch?
The MCP ecosystem is erupting, with thousands of MCP servers available for use. But developers face three major pain points:
| Pain points | Traditional approach | IntentOrch approach |
|---|---|---|
| Discovering Difficulty | Need to manually search and read documents | Natural language description requirements, AI automatically matches the best tool |
| Compile and solve difficulties | Need to write code to concatenate multiple tools | Generate a complete workflow in one sentence, automatically handle dependencies |
| Management is difficult | Each server needs to be started and monitored separately | Unified lifecycle management, visualized through web console operation |
IntentOrch transforms MCP from a 'developer tool' to a 'helper for everyone'.
______________________________________________________________________
✨ Core Features
🧠 Intent driven orchestration
你输入： "帮我查一下明天从广州到长沙的高铁票"
IntentOrch 自动完成：
┌─────────────────────────────────────────────────────┐
│ ① 意图解析 │
│ └─ 识别需求：查询火车票 │
│ └─ 提取参数：明天、广州→长沙、高铁 │
│ │
│ ② 工具匹配 │
│ └─ 找到 12306-mcp → get-tickets 工具 │
│ └─ 参数映射：广州→GZQ, 长沙→CSQ, 高铁→G │
│ │
│ ③ 工作流编排 │
│ └─ 生成可执行的工作流（1个步骤） │
│ └─ 自动处理依赖关系 │
│ │
│ ④ 自动执行 │
│ └─ 启动 MCP Server → 调用工具 → 返回结果 │
└─────────────────────────────────────────────────────┘No need to write code, no need to check documentation, just speak in person.
🐳 MCP Server Lifecycle Management
Manage your MCP Server like Docker:
| Command | Description | Analogous to Docker |
|---|---|---|
intorch pull | Retrieve MCP Server from the registry | docker pull |
intorch start | Start MCP Server | docker run |
intorch stop | Stop MCP Server | docker stop |
intorch ps | View running servers | docker ps |
intorch logs | View Server logs | docker logs |
support GitHubTheGitee Wait for multiple registry sources and pull MCP Server contributed by the community with one click.
🖥️ Visual Web Management Console
| Page | Function |
|---|---|
| 📊 dashboard | Overview of System Status, Server Health Status at a Glance |
| 💬 Intentional arrangement | AI chat style interaction, natural language → automated workflow |
| 📦 MCP Service | Pull, start, stop, and monitor MCP Server |
| 📋 workflow | View, edit, and execute saved workflows |
| 🔧 process monitoring | Real time viewing of process resource usage and running status |
| ⚙️ Configuration Management | AI provider and registry configuration |
| 🔑 Key Management | Safely store sensitive information such as API keys |
| 📝 Log viewing | Centralized system viewing Server、 Workflow log |
左：工作流管理 ｜ 右：MCP服务管理
🔧 CLI+Daemon+Web three in one
# 1.安装
npm install -g @intentorch/cli && npm install -g @intentorch/web
# 2.配置大模型(以deepseek为例）
intorch config set provider deepseek
intorch config set model deepseek-chat
intorch config set apiKey sk-xxx（替换为你自己的apiKey）
# 3.启动daemon
intorch daemon start
# 4.拉取mcp服务
intorch pull Joooook/12306-mcp
# 5.启动mcp服务
intorch start Joooook/12306-mcp
# 6.执行自然语言驱动的任务
intorch run "查询明天广州到南宁的高铁票"
intorch dashboard（想要通过web端控制台执行任务时可通过此命令启动）______________________________________________________________________
🏗️ Architectural Design
┌─────────────────────────────────────────────────────────┐
│ 用户交互层 │
│ ┌──────────┐ ┌──────────┐ ┌──────────────────┐ │
│ │ CLI │ │ Daemon │ │ Web Dashboard │ │
│ │ 命令行 │ │ 守护进程 │ │ 可视化控制台 │ │
│ └────┬─────┘ └────┬─────┘ └────────┬─────────┘ │
└───────┼───────────────┼───────────────────┼──────────────┘
│ │ │
┌───────┴───────────────┴───────────────────┴──────────────┐
│ 核心引擎层 │
│ │
│ ┌────────────────────────────────────────────────────┐ │
│ │ 意图驱动编排引擎 (Intent Engine) │ │
│ │ │ │
│ │ 自然语言 → 意图解析 → 工具匹配 → 工作流编排 → 执行 │ │
│ └────────────────────────────────────────────────────┘ │
│ │
│ ┌──────────────┐ ┌──────────────┐ ┌───────────────┐ │
│ │ Workflow │ │ Process │ │ Runtime │ │
│ │ Engine │ │ Manager │ │ Adapter │ │
│ │ 工作流引擎 │ │ 进程管理 │ │ 运行时适配 │ │
│ └──────────────┘ └──────────────┘ └───────────────┘ │
│ │
│ ┌──────────────┐ ┌──────────────┐ ┌───────────────┐ │
│ │ Tool │ │ Secret │ │ Config │ │
│ │ Registry │ │ Manager │ │ Manager │ │
│ │ 工具注册表 │ │ 密钥管理 │ │ 配置管理 │ │
│ └──────────────┘ └──────────────┘ └───────────────┘ │
└──────────────────────────────────────────────────────────┘
│ │
┌───────┴────────────────────┴─────────────────────────────┐
│ MCP 生态层 │
│ ┌──────────┐ ┌──────────┐ ┌──────────────────────┐ │
│ │ 12306 │ │ File │ │ 1000+ MCP │ │
│ │ MCP │ │ System │ │ Servers... │ │
│ └──────────┘ └──────────┘ └──────────────────────┘ │
└──────────────────────────────────────────────────────────┘core module
| Module | Description |
|---|---|
| @intentorch/core | Core business logic: intent parsing, workflow engine, MCP client, process management |
| @intentorch/cli | Command line tools: Server management, workflow management, daemon process control |
| @intentorch/web | Web management console: React+TypeScript+Tailwind CSS |
______________________________________________________________________
🚦 Quick Start
Prerequisite
- Node.js >= 18.0.0
- pnpm >= 8.0.0
# 1.安装
npm install -g @intentorch/cli && npm install -g @intentorch/web
# 2.配置大模型(以deepseek为例）
intorch config set provider deepseek
intorch config set model deepseek-chat
intorch config set apiKey sk-xxx（替换为你自己的apiKey）
# 3.启动daemon
intorch daemon start
# 4.拉取mcp服务
intorch pull Joooook/12306-mcp
# 5.启动mcp服务
intorch start Joooook/12306-mcp
# 6.执行自然语言驱动的任务
intorch run "查询明天广州到南宁的高铁票"
intorch dashboard（想要通过web端控制台执行任务时可通过此命令启动）______________________________________________________________________
📸 Example scenario
Scenario 1: Checking train tickets in one sentence
你： "帮我查一下明天从广州到长沙的高铁票"
IntentOrch：
✅ 解析意图 → 查询火车票
✅ 匹配工具 → 12306-mcp / get-tickets
✅ 参数映射 → date=明天, from=广州, to=长沙, filter=高铁
✅ 执行结果 → 返回 G 字头列车时刻表和余票信息Scenario 2: Multi step automation
你： "每天早上8点检查服务器状态，如果 CPU 超过 80% 就发送告警"
IntentOrch：
✅ 解析为 3 个原子意图
├─ 步骤1: 获取服务器 CPU 使用率
├─ 步骤2: 判断是否超过阈值
└─ 步骤3: 发送告警通知
✅ 自动编排依赖关系
✅ 生成可定时执行的工作流Scenario 3: Combining multiple MCP servers
你： "搜索最新的 AI 新闻，总结要点，保存到本地文件"
IntentOrch：
✅ 步骤1: search-web → 搜索 AI 新闻
✅ 步骤2: ai-summarize → 总结要点
✅ 步骤3: filesystem → 保存到本地
✅ 自动处理步骤间的数据传递______________________________________________________________________
🧩 Ecological compatibility
IntentOrch is compatible with all following Model Context Protocol Standard MCP Server:
- 📊 database - PostgreSQL, MySQL, SQLite queries
- 📁 file system - File read/write, directory operations
- 🌐 web service - HTTP requests, API calls
- 🚄 Transportation 12306 ticket and flight inquiry
- 📰 Information acquisition - News, weather, encyclopedia
- 🛠️ development tools Git operations, code analysis
______________________________________________________________________
📄 license
This project is based on Apache 2.0 License open source.
______________________________________________________________________
IntentOrch — 从"告诉计算机怎么做"到"告诉计算机你想要什么"
GitHub • Gitee • Issues