speckit-plus-mcp 0.1.0

MCP server for Spec-Kit Plus — connect any AI agent to spec-driven development

🛠️ Spec-Kit Plus MCP Server

MCP (Model Context Protocol) server that wraps Spec-Kit Plus tools so any AI IDE or CLI (Qwen, Gemini, Claude Code, Cursor, etc.) can use it globally.

📦 What This Is

This MCP server acts as a bridge between AI agents and the Spec-Kit Plus CLI (specifyplus), enabling spec-driven development workflows across any MCP-compatible AI tool.

User (in any AI IDE/CLI)
        │
        ▼
   MCP Protocol
        │
        ▼
  speckit-plus-mcp/
    server.py          ← MCP Server
        │
        ▼
  specifyplus CLI      ← Spec-Kit Plus (sp init, sp.plan, sp.tasks, etc.)

✅ Prerequisites

# Install specifyplus (the CLI being wrapped)
pip install specifyplus

# Install MCP Python SDK
pip install "mcp[cli]"

🚀 Installation

1. Clone or download this repository

2. Install dependencies:

   cd speckit-plus-mcp
   pip install -e .
   

3. Configure your AI agent (see below)

🔧 Configuration

Qwen CLI

Create ~/.qwen/settings.json:

{
  "mcpServers": {
    "speckit-plus": {
      "command": "python",
      "args": ["/ABSOLUTE/PATH/TO/speckit-plus-mcp/server.py"],
      "env": {
        "PYTHONPATH": "."
      }
    }
  }
}

⚠️ Replace /ABSOLUTE/PATH/TO/ with your actual path.

Gemini CLI

Create ~/.gemini/settings.json:

{
  "mcpServers": {
    "speckit-plus": {
      "command": "python",
      "args": ["/ABSOLUTE/PATH/TO/speckit-plus-mcp/server.py"]
    }
  }
}

Claude Code

Create ~/.claude/claude_desktop_config.json:

{
  "mcpServers": {
    "speckit-plus": {
      "command": "python",
      "args": ["/ABSOLUTE/PATH/TO/speckit-plus-mcp/server.py"]
    }
  }
}

Cursor / VS Code (with MCP extension)

Add to .cursor/mcp.json inside your project:

{
  "mcpServers": {
    "speckit-plus": {
      "command": "python",
      "args": ["/ABSOLUTE/PATH/TO/speckit-plus-mcp/server.py"]
    }
  }
}

🧪 Testing

Test the server runs correctly:

cd speckit-plus-mcp
python server.py

You should see: Running MCP server 'speckit-plus'

Test inside Qwen CLI:

qwen

Then type /mcp — you should see:

speckit-plus: CONNECTED
Tools: speckit_init, speckit_check, speckit_specify, speckit_plan ...

📋 Available Tools

Tool	Description
speckit_init	Initialize a new Spec-Kit Plus project for any AI agent
speckit_check	Verify all required tools are installed
speckit_constitution	Set project governing principles
speckit_specify	Write requirements specification (WHAT to build)
speckit_clarify	Run structured clarification before planning
speckit_plan	Generate technical plan with tech stack (HOW to build)
speckit_tasks	Break plan into ordered task list
speckit_implement	Execute all tasks to build the full project
speckit_workflow_guide	Get complete step-by-step workflow instructions

📘 Usage Example

In your AI agent (Qwen, Claude, etc.):

Use speckit_init to create a project called "todo-app" for qwen,
then show me the workflow guide.

Complete Workflow:

1. Initialize project:

   speckit_init(project_name="my-app", ai_agent="qwen")
   

2. Set principles:

   /sp.constitution Enforce clean code, testing, and performance
   

3. Write specification:

   /sp.specify Build a task manager with Kanban boards and user assignments
   

4. Clarify gaps (recommended):

   /sp.clarify
   

5. Technical plan:

   /sp.plan Use FastAPI, PostgreSQL, React, Docker
   

6. Generate tasks:

   /sp.tasks
   

7. Build the project:

   /sp.implement
   

🎯 Supported AI Agents

• qwen
• claude
• gemini
• copilot
• cursor
• windsurf
• opencode
• codex
• amp

📁 Output Files

After running the workflow, you'll find these files in .specify/:

.specify/
├── memory/
│   └── constitution.md    ← Your project principles
└── specs/
    └── 001-feature/
        ├── spec.md        ← Requirements & user stories
        ├── plan.md        ← Technical implementation plan
        ├── tasks.md       ← Ordered task list
        └── data-model.md  ← Database schema

🏗️ How It Works

AI Agent (Qwen/Gemini/Claude)
        │
        │  speaks MCP protocol
        ▼
   server.py  (MCP server)
        │
        │  calls subprocess
        ▼
   specifyplus CLI  (sp commands)
        │
        │  reads/writes
        ▼
   .specify/ folder
   (spec.md, plan.md, tasks.md)
        │
        │  used by
        ▼
   AI Agent builds the project

MCP = Model Context Protocol — a standard that lets any AI agent talk to any tool server. This server.py IS the bridge between the AI and Spec-Kit Plus.

📄 License

MIT