
Spec Workflow MCP
MCP规格驱动开发 MCP 服务器:让 agent 按需求、设计、任务三步写规格文档并逐项实现,配有网页看板和 VS Code 扩展,产品经理可以在看板里审阅、反馈、批准文档并跟踪任务进度。需要 Node.js;网页看板需用 npx 另行启动,VS Code 用户可改装扩展。
Spec Workflow MCP
I HAVE TAKEN A SMALL BREAK FROM THIS REPO FOR PERSONAL REASONS BUT I WILL BE BACK WITH SOME UPDATES IN THE NEAR FUTURE THANK YOU FOR YOUR UNDERSTANDING
A Model Context Protocol (MCP) server for structured spec-driven development with real-time dashboard and VSCode extension.
☕ Support This Project
📺 Showcase
🔄 Approval System in Action
See how the approval system works: create documents, request approval through the dashboard, provide feedback, and track revisions.
📊 Dashboard & Spec Management
Explore the real-time dashboard: view specs, track progress, navigate documents, and monitor your development workflow.
✨ Key Features
- Structured Development Workflow - Sequential spec creation (Requirements → Design → Tasks)
- Real-Time Web Dashboard - Monitor specs, tasks, and progress with live updates
- VSCode Extension - Integrated sidebar dashboard for VSCode users
- Approval Workflow - Complete approval process with revisions
- Task Progress Tracking - Visual progress bars and detailed status
- Implementation Logs - Searchable logs of all task implementations with code statistics
- Multi-Language Support - Available in 11 languages
🌍 Supported Languages
🇺🇸 English • 🇯🇵 日本語 • 🇨🇳 中文 • 🇪🇸 Español • 🇧🇷 Português • 🇩🇪 Deutsch • 🇫🇷 Français • 🇷🇺 Русский • 🇮🇹 Italiano • 🇰🇷 한국어 • 🇸🇦 العربية
📖 Documentation in your language:
English | 日本語 | 中文 | Español | Português | Deutsch | Français | Русский | Italiano | 한국어 | العربية
🚀 Quick Start
Step 1: Add to your AI tool
Add to your MCP configuration (see client-specific setup below):
{
"mcpServers": {
"spec-workflow": {
"command": "npx",
"args": ["-y", "@pimzino/spec-workflow-mcp@latest", "/path/to/your/project"]
}
}
}
Step 2: Choose your interface
Option A: Web Dashboard (Required for CLI users) Start the dashboard (runs on port 5000 by default):
npx -y @pimzino/spec-workflow-mcp@latest --dashboard
The dashboard will be accessible at: http://localhost:5000
Note: Only one dashboard instance is needed. All your projects will connect to the same dashboard.
Option B: VSCode Extension (Recommended for VSCode users)
Install Spec Workflow MCP Extension from the VSCode marketplace.
📝 How to Use
Simply mention spec-workflow in your conversation:
- "Create a spec for user authentication" - Creates complete spec workflow
- "List my specs" - Shows all specs and their status
- "Execute task 1.2 in spec user-auth" - Runs a specific task
🔧 MCP Client Setup
Augment Code
Configure in your Augment settings:
{
"mcpServers": {
"spec-workflow": {
"command": "npx",
"args": ["-y", "@pimzino/spec-workflow-mcp@latest", "/path/to/your/project"]
}
}
}
Claude Code CLI
Add to your MCP configuration:
claude mcp add spec-workflow npx @pimzino/spec-workflow-mcp@latest -- /path/to/your/project
Important Notes:
- The
-yflag bypasses npm prompts for smoother installation - The
--separator ensures the path is passed to the spec-workflow script, not to npx - Replace
/path/to/your/projectwith your actual project directory path
Alternative for Windows (if the above doesn't work):
claude mcp add spec-workflow cmd.exe /c "npx @pimzino/spec-workflow-mcp@latest /path/to/your/project"
Claude Desktop
Add to claude_desktop_config.json:
{
"mcpServers": {
"spec-workflow": {
"command": "npx",
"args": ["-y", "@pimzino/spec-workflow-mcp@latest", "/path/to/your/project"]
}
}
}
Important: Run the dashboard separately with
--dashboardbefore starting the MCP server.
Cline/Claude Dev
Add to your MCP server configuration:
{
"mcpServers": {
"spec-workflow": {
"command": "npx",
"args": ["-y", "@pimzino/spec-workflow-mcp@latest", "/path/to/your/project"]
}
}
}
Continue IDE Extension
Add to your Continue configuration:
{
"mcpServers": {
"spec-workflow": {
"command": "npx",
"args": ["-y", "@pimzino/spec-workflow-mcp@latest", "/path/to/your/project"]
}
}
}
Cursor IDE
Add to your Cursor settings (settings.json):
{
"mcpServers": {
"spec-workflow": {
"command": "npx",
"args": ["-y", "@pimzino/spec-workflow-mcp@latest", "/path/to/your/project"]
}
}
}
OpenCode
Add to your opencode.json configuration file:
{
"$schema": "https://opencode.ai/config.json",
"mcp": {
"spec-workflow": {
"type": "local",
"command": ["npx", "-y", "@pimzino/spec-workflow-mcp@latest", "/path/to/your/project"],
"enabled": true
}
}
}
Windsurf
Add to your ~/.codeium/windsurf/mcp_config.json configuration file:
{
"mcpServers": {
"spec-workflow": {
"command": "npx",
"args": ["-y", "@pimzino/spec-workflow-mcp@latest", "/path/to/your/project"]
}
}
}
Codex
Add to your ~/.codex/config.toml configuration file:
[mcp_servers.spec-workflow]
command = "npx"
args = ["-y", "@pimzino/spec-workflow-mcp@latest", "/path/to/your/project"]
🐳 Docker Deployment
Run the dashboard in a Docker container for isolated deployment:
# Using Docker Compose (recommended)
cd containers
docker-compose up --build
# Or using Docker CLI
docker build -f containers/Dockerfile -t spec-workflow-mcp .
docker run -p 5000:5000 -v "./workspace/.spec-workflow:/workspace/.spec-workflow:rw" spec-workflow-mcp
The dashboard will be available at: http://localhost:5000
🔒 Security
Spec-Workflow MCP includes enterprise-grade security features suitable for corporate environments:
✅ Implemented Security Controls
| Feature | Description |
|---|---|
| Localhost Binding | Binds to 127.0.0.1 by default, preventing network exposure |
| Rate Limiting | 120 requests/minute per client with automatic cleanup |
| Audit Logging | Structured JSON logs with timestamp, actor, action, and result |
| Security Headers | X-Content-Type-Options, X-Frame-Options, X-XSS-Protection, CSP, Referrer-Policy |
| CORS Protection | Restricted to localhost origins by default |
| Docker Hardening | Non-root user, read-only filesystem, dropped capabilities, resource limits |
⚠️ Not Yet Implemented
| Feature | Workaround |
|---|---|
| HTTPS/TLS | Use a reverse proxy (nginx, Apache) with TLS certificates |
| User Authentication | Use a reverse proxy with Basic Auth or OAuth2 Proxy for SSO |
For External/Network Access
If you need to expose the dashboard beyond localhost, we recommend:
- Keep dashboard on localhost (
127.0.0.1) - Use nginx or Apache as a reverse proxy with:
- TLS/HTTPS termination
- Basic authentication or OAuth2
- Configure firewall rules to restrict access
# Example nginx reverse proxy with auth
server {
listen 443 ssl;
server_name dashboard.example.com;
ssl_certificate /path/to/cert.pem;
ssl_certificate_key /path/to/key.pem;
auth_basic "Dashboard Access";
auth_basic_user_file /etc/nginx/.htpasswd;
location / {
proxy_pass http://127.0.0.1:5000;
proxy_http_version 1.1;
proxy_set_header Upgrade $http_upgrade;
proxy_set_header Connection "upgrade";
}
}
🔒 Sandboxed Environments
For sandboxed environments (e.g., Codex CLI with sandbox_mode=workspace-write) where $HOME is read-only, use the SPEC_WORKFLOW_HOME environment variable to redirect global state files to a writable location:
SPEC_WORKFLOW_HOME=/workspace/.spec-workflow-mcp npx -y @pimzino/spec-workflow-mcp@latest /workspace
📚 Documentation
- Configuration Guide - Command-line options, config files
- User Guide - Comprehensive usage examples
- Workflow Process - Development workflow and best practices
- Interfaces Guide - Dashboard and VSCode extension details
- Prompting Guide - Advanced prompting examples
- Tools Reference - Complete tools documentation
- Development - Contributing and development setup
- Troubleshooting - Common issues and solutions
📁 Project Structure
your-project/
.spec-workflow/
approvals/
archive/
specs/
steering/
templates/
user-templates/
config.example.toml
🛠️ Development
# Install dependencies
npm install
# Build the project
npm run build
# Run in development mode
npm run dev
📄 License
GPL-3.0
⭐ Star History
相關資源

Task Master
MCP把 PRD 解析成带依赖关系的任务清单,让 Cursor、Claude Code 等编码 agent 逐项拆解、实现并跟踪进度,也可当 CLI 使用。需配置至少一个模型 API Key,或复用 Claude Code / Codex CLI 登录。
eyaltoledano/claude-task-master28.5k
BMad Method
Skill敏捷 AI 驱动开发方法:内置分析师、产品经理、UX 设计师等角色技能,把想法整理成产品简报、PRD 与 PRFAQ,再拆成任务交给 agent 实现。需 Claude Code 等支持技能的工具,并装好 Node.js、Git 与 uv,之后让 bmad 技能执行 bmad setup。
bmad-code-org/BMAD-METHOD55.6k
Taskmaster
Plugin把需求文档(PRD)解析成带依赖关系的任务列表,支持复杂度分析、任务拆解与调研;插件打包了斜杠命令、子代理与 MCP 服务。需配置模型 API Key,或复用 Claude Code / Codex CLI 登录。
eyaltoledano/claude-task-master28.5k
Task Master
CLI把 PRD 解析成带依赖关系的任务清单,让 Claude Code、Cursor 等 agent 按序领取下一项、拆分子任务、更新进度,同时提供 CLI 与 MCP。需要至少一个模型 API Key,也可复用 Claude Code / Codex 登录。
eyaltoledano/claude-task-master28.4k
CCPM
Skill软件项目的规格驱动管理流程:先引导头脑风暴写出 PRD,再拆成 Epic 与带依赖的任务并同步为 GitHub Issues,可随时询问站会进度、下一步与阻塞项,也能派多个 agent 并行写代码。需要 Git、已登录的 gh 命令行和一个 GitHub 仓库。
automazeio/ccpm8.5k
notion-spec-to-implementation
Skill从 Notion 读取 PRD 或功能规格,梳理需求与待澄清点,生成实施计划页并拆出任务,把规格、计划与任务互相关联并持续更新进度。收录于 OpenAI 官方插件仓库(开发者标注为 Notion),需要连接 Notion MCP。
openai/plugins8k
