Spec Kit MCP
MCP server enabling AI assistants to use GitHub's spec-kit methodology
Ask AI about Spec Kit MCP
Powered by Claude Β· Grounded in docs
I know everything about Spec Kit MCP. Ask me about installation, configuration, usage, or troubleshooting.
0/500
Reviews
Documentation
Spec-Kit MCP Server
MCP server that enables AI coding assistants to use spec-driven development practices via the GitHub Spec-Kit toolkit.
Features
- π― 100% Spec-Kit Coverage: All 10 tools for complete spec-driven development
- π MCP Protocol: Full JSON-RPC 2.0 implementation for AI agents
- β‘ High Performance: Built with Rust and Tokio for async I/O
- π§ Dual Installation: Install via
cargoornpx - π‘οΈ Type Safe: Comprehensive type system with validation
- π Comprehensive Docs: Tutorials, examples, and configuration guides
- π Multi-Editor Support: Claude Code, Cursor, Windsurf, VS Code, and more
- π§ͺ Production Ready: Tested, documented, and deployed
Quick Start
Installation
Via Cargo (Recommended)
Fast, reliable, works offline:
cargo install spec-kit-mcp
Advantages:
- β Fastest startup
- β Works offline
- β Most reliable
- β Full platform support (macOS Intel/ARM, Linux)
Via npm/npx
For Node.js users:
# Global installation
npm install -g @lsendel/spec-kit-mcp
# Or use with npx (downloads on first use)
npx @lsendel/spec-kit-mcp
Advantages:
- β Familiar for Node.js users
- β Auto-downloads prebuilt binaries
- β npx always uses latest version
Prerequisites
- Python 3.11+: Required by GitHub spec-kit
- uv package manager: Required to run spec-kit (install from https://docs.astral.sh/uv/)
- GitHub Spec-Kit: No separate installation needed - the MCP server uses
uvxto run spec-kit directly - Node.js 18+: Only if using npx installation method
- Git: For version control operations
Note: The spec-kit CLI is not available on PyPI. The MCP server automatically runs it via uvx --from git+https://github.com/github/spec-kit.git
Configuration with Claude Code
Claude Code supports two ways to configure MCP servers:
Method 1: Using Cargo Binary (Recommended)
First install via cargo:
cargo install spec-kit-mcp
Then configure ~/.config/claude-code/mcp.json:
{
"mcpServers": {
"spec-kit": {
"command": "spec-kit-mcp",
"args": [],
"env": {}
}
}
}
Advantages:
- β Fastest startup (<100ms)
- β Works offline
- β Most reliable
- β Supports all platforms (macOS Intel/ARM, Linux)
Method 2: Using npx (No Installation)
Create or edit ~/.config/claude-code/mcp.json:
{
"mcpServers": {
"spec-kit": {
"command": "npx",
"args": ["-y", "@lsendel/spec-kit-mcp"],
"env": {}
}
}
}
Advantages:
- β No installation required
- β Always uses latest version
- β Good for trying it out
Note: First run may be slower as it downloads the binary.
Verify Configuration
-
Restart Claude Code after editing the config file
-
Test the MCP tools in Claude Code:
List all available MCP tools
You should see 10 spec-kit tools listed:
- speckit_init
- speckit_check
- speckit_constitution
- speckit_specify
- speckit_plan
- speckit_tasks
- speckit_implement
- speckit_clarify
- speckit_analyze
- speckit_checklist
- Try a simple command:
Use speckit_check to verify my development environment
Troubleshooting Configuration
Issue: Tools not appearing after restart
Solution:
- Check config file location:
cat ~/.config/claude-code/mcp.json - Verify JSON syntax:
cat ~/.config/claude-code/mcp.json | jq - Check Claude Code logs:
~/.config/claude-code/logs/ - For npx method, ensure Node.js is installed:
node --version - For binary method, verify installation:
which spec-kit-mcp
Issue: spec-kit-mcp: command not found (Method 2)
Solution:
- Use Method 1 (npx) instead, or
- Add npm global bin to PATH:
export PATH="$PATH:$(npm config get prefix)/bin"
Need help with other editors? See the Complete Configuration Guide for Cursor, Windsurf, VS Code, and more.
π Documentation
Quick Links
- π Comprehensive Tutorials - Step-by-step guides for all skill levels
- π‘ Examples - Real-world project examples
- βοΈ Configuration Guide - Setup for different AI assistants
- π Usage Guide - Detailed usage documentation
Learning Resources
New to Spec-Kit? Start here:
- Tutorial 1: Your First Spec-Kit Project (20 min)
- Todo CLI Example - Complete beginner example
- Configuration for Your Editor
Building APIs? Check out:
- Tutorial 2: Building a REST API
- Blog API Example - Production-ready API example
Working with teams?
Available Tools
The MCP server exposes 10 spec-kit tools for the complete workflow:
1. speckit_init
Initialize a new spec-kit project with proper structure.
{
"project_name": "my-project",
"project_path": "."
}
2. speckit_constitution
Create project governing principles and development standards.
{
"principles": "Simplicity, Performance, Security",
"constraints": "Must support Python 3.11+",
"output_path": "./speckit.constitution"
}
3. speckit_specify
Define requirements and user stories (the "what").
{
"requirements": "User authentication system with OAuth2 support",
"user_stories": "As a user, I want to login with Google...",
"output_path": "./speckit.specify"
}
4. speckit_plan
Create a technical implementation plan (the "how").
{
"spec_file": "./speckit.specify",
"tech_stack": "Rust + Tokio",
"output_path": "./speckit.plan"
}
5. speckit_tasks
Generate actionable task lists from the plan.
{
"plan_file": "./speckit.plan",
"breakdown_level": "medium",
"output_path": "./speckit.tasks"
}
6. speckit_implement
Execute implementation according to the task list.
{
"task_file": "./speckit.tasks",
"context": "Using Rust with async/await",
"output_dir": "./src"
}
7. speckit_clarify
Request clarification on ambiguous requirements or specifications.
{
"spec_file": "./speckit.specify",
"questions": "How should we handle edge cases?"
}
8. speckit_analyze
Analyze code for quality, compliance, and technical debt.
{
"target_path": "./src",
"check_constitution": true,
"output_format": "markdown"
}
9. speckit_check
Validate that required tools are installed for spec-kit development.
{
"check_speckit": true,
"check_git": true,
"check_ai_tools": true
}
10. speckit_checklist
Generate review checklists to verify implementation completeness.
{
"spec_file": "./speckit.specify",
"task_file": "./speckit.tasks",
"output_path": "./checklist.md"
}
See all tools in action: Check out the Examples directory for complete workflows
Usage Example
Here's a complete workflow using Claude Code:
User: Initialize a new spec-kit project called "user-auth"
Claude: [Uses speckit_init tool]
β Project initialized at ./user-auth
User: Create a constitution focusing on security and simplicity
Claude: [Uses speckit_constitution tool]
β Constitution created at ./speckit.constitution
User: Specify requirements for OAuth2 authentication
Claude: [Uses speckit_specify tool]
β Specification created at ./speckit.specify
User: Create a technical plan using Rust and OAuth2 libraries
Claude: [Uses speckit_plan tool]
β Technical plan created at ./speckit.plan
User: Generate detailed tasks
Claude: [Uses speckit_tasks tool]
β Task list created at ./speckit.tasks
Found 15 actionable tasks
Architecture
AI Agent (Claude Code, Cursor, etc.)
β
MCP Protocol (JSON-RPC 2.0 over stdio)
β
Spec-Kit MCP Server (Rust/Tokio)
β
Tool Registry & Dispatcher
β
Spec-Kit CLI Integration Layer
β
Spec-Kit Python CLI (subprocess)
β
File System (speckit.* artifacts)
Development
Building from Source
git clone https://github.com/yourusername/spec-kit-mcp.git
cd spec-kit-mcp
cargo build --release
Running Tests
cargo test
Running the Server
# With default settings
cargo run
# With custom log level
cargo run -- --log-level debug
# With custom CLI path
cargo run -- --cli-path /path/to/specify
# With custom timeout
cargo run -- --timeout 600
Project Structure
spec-kit-mcp/
βββ src/
β βββ main.rs # Binary entry point
β βββ lib.rs # Library root
β βββ mcp/ # MCP protocol implementation
β β βββ types.rs # JSON-RPC types
β β βββ protocol.rs # Protocol handler
β β βββ transport.rs # Stdio transport
β β βββ server.rs # MCP server
β βββ speckit/ # Spec-kit CLI integration
β β βββ cli.rs # Command execution
β β βββ errors.rs # Error types
β βββ tools/ # MCP tools
β βββ mod.rs # Tool registry
β βββ init.rs # speckit_init tool
β βββ constitution.rs # speckit_constitution tool
β βββ specify.rs # speckit_specify tool
β βββ plan.rs # speckit_plan tool
β βββ tasks.rs # speckit_tasks tool
βββ Cargo.toml # Rust package manifest
βββ README.md # This file
Contributing
Contributions are welcome! Please see CONTRIBUTING.md for guidelines.
Development Workflow
- Fork the repository
- Create a feature branch:
git checkout -b feature/my-feature - Make your changes
- Run tests:
cargo test - Run clippy:
cargo clippy - Format code:
cargo fmt - Commit with conventional commits:
git commit -m "feat: add new feature" - Push and create a Pull Request
Roadmap
Current Version (0.1.0)
- β All 10 spec-kit tools implemented (100% coverage)
- β MCP protocol support (JSON-RPC 2.0)
- β Dual distribution (cargo + npx)
- β Comprehensive error handling
- β Complete tutorials and examples
- β Configuration guides for major editors
- β Published to crates.io and npm
Future Versions
v0.2.0
- Enhanced tool parameters and validation
- Configuration file support (.speckit-mcp.toml)
- Template system for common project types
- Performance optimizations and caching
- Windows platform support
- Web-based tool output visualization
v0.3.0
- Remote MCP via Server-Sent Events (SSE)
- Web UI dashboard for project visualization
- Template marketplace integration
- Team collaboration features (shared constitutions)
- Metrics and analytics dashboard
- Plugin system for custom tools
Performance
- Cold start: <500ms
- Tool invocation: <200ms (excluding spec-kit CLI execution)
- Memory usage: <50MB baseline
- Concurrent requests: 10+
Compatibility
AI Coding Assistants
- β Claude Code
- β Cursor
- β GitHub Copilot (with MCP support)
- β Any MCP-compatible client
Platforms
- β macOS (Intel and ARM)
- β Linux (x86_64)
- β³ Windows (planned)
Troubleshooting
Spec-Kit CLI Not Found
Error: spec-kit CLI not found!
Solution: Ensure uv package manager is installed:
# Check if uv is installed
uv --version
# If not installed, install uv (macOS/Linux)
curl -LsSf https://astral.sh/uv/install.sh | sh
# Or using pip
pip install uv
# Test spec-kit access
uvx --from git+https://github.com/github/spec-kit.git specify check
Note: The spec-kit CLI is not available as a standalone package. The MCP server uses uvx to run it directly from GitHub.
Python Version Too Old
Solution: Upgrade to Python 3.11 or later:
# Check version
python3 --version
# Install Python 3.11+ (macOS with Homebrew)
brew install python@3.11
Permission Denied
Solution: Ensure the binary is executable:
chmod +x $(which spec-kit-mcp)
FAQ
Q: Do I need to install spec-kit separately? A: Yes, the MCP server requires the spec-kit Python CLI to be installed.
Q: Can I use this without Claude Code? A: Yes! It works with any MCP-compatible AI coding assistant.
Q: Does this work offline? A: Yes, if installed via cargo. The npx version requires internet for initial download.
Q: How does this compare to using spec-kit directly? A: This MCP server enables AI agents to use spec-kit automatically, streamlining the workflow.
License
This project is dual-licensed under:
- MIT License (LICENSE-MIT or http://opensource.org/licenses/MIT)
- Apache License 2.0 (LICENSE-APACHE or http://www.apache.org/licenses/LICENSE-2.0)
You may choose either license for your use.
Acknowledgments
- GitHub Spec-Kit - The underlying spec-driven development toolkit
- Model Context Protocol - The protocol enabling AI-tool integration
- Anthropic Claude - AI assistant that inspired this integration
- Rust Community - For excellent async tooling (Tokio, Serde, etc.)
Links
- Crates.io: https://crates.io/crates/spec-kit-mcp
- Docs.rs: https://docs.rs/spec-kit-mcp
- GitHub: https://github.com/yourusername/spec-kit-mcp
- Issues: https://github.com/yourusername/spec-kit-mcp/issues
- Spec-Kit: https://github.com/github/spec-kit
Built with Rust | Production Ready | Open Source
