About shotflow-mcp
shotflow-mcp exposes ShotFlow storyboard and video-generation capabilities to MCP clients, including tools for tracking and revising video tasks. It also provides resources for creative types and visual styles.
Key features
- 01Generate storyboards from scripts or ideas
- 02Generate videos from storyboard shots
- 03Check video-generation status
- 04Regenerate individual shots
- 05List video task history and check quota
- 06Access creative-type and visual-style resources
Use cases
- 01Turning a script idea into a storyboard
- 02Creating a video from storyboard shots
- 03Checking progress or revising a generated shot
- 04Reviewing task history and available creative templates
MCP support
It provides seven tools and two template resources to MCP clients over Streamable HTTP. An online endpoint is available at `https://aishotflow.cn/mcp`.
Configuration
Local server config is the same for every client. Copy it into your client’s MCP config file.
npx -y shotflow-mcp-serverSHOTFLOW_API_URLShotFlow Express API 地址(URL 格式,例如 http://localhost:3003)MCP_API_KEYSecretMCP Server 认证 Key(可选)Local servers need a runtime on your machine (e.g. Node.js 18+). Config file locations differ by client; see your client’s docs.
Tools
Local servers expose tools only once running on your machine, and the maintainer has not provided a tool list.
Details
From the project READMEShotFlow MCP Server
让 AI Agent 直接调用专业分镜脚本生成能力
ShotFlow MCP Server 是 ShotFlow 的 Model Context Protocol 接口实现,让任何支持 MCP 的 AI 客户端(Claude Desktop、Cursor、Cline、VS Code Copilot 等)都能直接调用专业分镜脚本与视频生成能力。
在线使用
无需安装,直接配置 URL 即可使用:
https://aishotflow.cn/mcp
功能一览
| Tool | 说明 |
|---|---|
generate_storyboard | 从剧本/创意描述生成专业分镜脚本(6种创作类型 × 7种视觉风格) |
generate_video | 从分镜镜头生成视频(Pro 功能) |
check_video_status | 查询视频生成状态 |
regenerate_shot | 单镜头重新生成 |
list_video_tasks | 列出视频任务历史 |
check_quota | 查询今日配额 |
check_health | 检查服务健康状态 |
| Resource | 说明 |
|---|---|
shotflow://templates/creative-types | 6种创作类型模板与默认参数 |
shotflow://templates/visual-styles | 7种视觉风格参考与关键词 |
快速接入
Claude Desktop
编辑配置文件(macOS: ~/Library/Application Support/Claude/claude_desktop_config.json,Windows: %APPDATA%\Claude\claude_desktop_config.json):
{
"mcpServers": {
"shotflow": {
"url": "https://aishotflow.cn/mcp"
}
}
}
重启 Claude Desktop,点击「Attached tools」确认 shotflow 已加载。
Cursor IDE
在项目根目录创建 .cursor/mcp.json:
{
"mcpServers": {
"shotflow": {
"url": "https://aishotflow.cn/mcp"
}
}
}
VS Code(1.99+)
在项目根目录创建 .vscode/mcp.json:
{
"servers": {
"shotflow": {
"type": "http",
"url": "https://aishotflow.cn/mcp"
}
}
}
Cline / Roo-Cline
编辑 ~/.cline/mcp_settings.json:
{
"mcpServers": {
"shotflow": {
"url": "https://aishotflow.cn/mcp"
}
}
}
调用示例
生成短视频分镜
告诉你的 AI 客户端:
用 shotflow 的 generate_storyboard 工具,帮我生成一个短视频分镜:一个创业者熬夜做产品的故事,3个场景,温情风格,9:16竖屏
AI 会自动调用 MCP Tool,返回结构化分镜结果。
查看可用创作类型
读取 shotflow 的 creative-types 资源,告诉我有哪些创作类型可选
Agent Card(A2A 协议)
其他 AI Agent 可通过以下 URL 自动发现 ShotFlow 的能力:
https://aishotflow.cn/mcp/agent-card
部署自己的实例
要求
- Node.js 18+
- ShotFlow Express Server 运行在
localhost:3003
安装
git clone https://github.com/YOUR_USERNAME/shotflow-mcp.git
cd shotflow-mcp
# 无需 npm install,零依赖
启动
# 开发模式
node shotflow-mcp-server.js
# 默认端口 3004
# 生产部署(systemd)
sudo cp shotflow-mcp.service /etc/systemd/system/
sudo systemctl enable shotflow-mcp
sudo systemctl start shotflow-mcp
环境变量
| 变量 | 默认值 | 说明 |
|---|---|---|
MCP_PORT | 3004 | MCP Server 监听端口 |
SHOTFLOW_API_URL | http://localhost:3003 | ShotFlow Express 地址 |
NODE_ENV | development | 环境模式 |
架构
AI Client (Claude/Cursor/...)
↓ Streamable HTTP (JSON-RPC 2.0)
Nginx :443
└── /mcp → ShotFlow MCP Server :3004
└── HTTP → ShotFlow Express :3003
- 协议: MCP 1.0(Streamable HTTP + SSE)
- 传输: JSON-RPC 2.0 over HTTP
- 依赖: 零外部依赖(纯 Node.js 实现)
- SSE 支持:
proxy_buffering off配合 Nginx 实现流式进度推送
文件结构
shotflow-mcp/
├── shotflow-mcp-server.js # MCP Server 主文件
├── tool-definitions.js # 7个 Tool 的 JSON Schema 定义
├── resource-definitions.js # 2个 Resource 定义
├── prompt-builder.js # Prompt 构造与模型配置解析
├── package.json # npm 包配置
├── shotflow-mcp.service # systemd 服务文件
├── INTEGRATION.md # 详细接入文档
├── README.md # 本文件
└── test/
└── test-mcp.js # 测试套件
测试
# 启动本地服务
node shotflow-mcp-server.js &
# 运行测试
node test/test-mcp.js
相关链接
- ShotFlow 主站: https://aishotflow.cn
- MCP 端点: https://aishotflow.cn/mcp
- Agent Card: https://aishotflow.cn/mcp/agent-card
- MCP 协议规范: https://modelcontextprotocol.io
- ShotFlow 公众号: 扫码关注获取最新更新
许可证
MIT License — 自由使用、修改和分发。
支持
- GitHub Issues: https://github.com/YOUR_USERNAME/shotflow-mcp/issues
- 邮件: [email protected]
Comments
New to MCP?
From what the protocol is to installing your first server, these answers cover most first questions.
An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.