The Complete AI-Enhanced Development Framework
Build better software with specifications first, powered by AI assistants across 8 major platforms
SpecPulse is not just another CLI tool - it's a complete development framework that bridges the gap between your ideas and production code. By combining structured specifications with AI-powered assistance, SpecPulse transforms how you build software.
Traditional development is chaotic:
- β Ideas jump directly to coding without proper planning
- β Requirements get lost in scattered documentation
- β AI assistants lack context about your project structure
- β Teams work without clear specifications or task breakdowns
SpecPulse introduces AI-Enhanced Specification-Driven Development:
- Specifications First - Clear, detailed specs before any code
- AI-Powered - Your favorite AI assistant helps create and expand specs
- Structured Planning - Break down features into manageable tasks
- Universal Integration - Works across 8 major AI platforms
- CLI Foundation - Reliable command-line structure with AI enhancement
SpecPulse works everywhere you work - supporting 8 major AI platforms with identical functionality:
| Platform | Integration Type | Commands | Status |
|---|---|---|---|
| Claude Code | Custom Slash Commands | /sp-* |
β Full Support |
| Gemini CLI | Custom Commands | /sp-* |
β Full Support |
| Windsurf | Custom Slash Commands | /sp-* |
β Full Support |
| Cursor | Custom Slash Commands | /sp-* |
β Full Support |
| GitHub Copilot | Custom Prompts | .prompt.md |
β Full Support |
| OpenCode | Agent-Based Workflow | /sp-* |
β Full Support |
| Crush | Category Commands | /sp-* |
β Full Support |
| Qwen Code | TOML Configuration | /sp-* |
β Full Support |
The same 11 commands work everywhere with identical functionality:
# These commands work identically on ALL 8 platforms:
/sp-pulse payment-system # Initialize new feature - START HERE
/sp-spec "User authentication" # Create specification
/sp-plan # Generate implementation plan
/sp-task # Break down into tasks
/sp-execute # Execute tasks continuously
/sp-status # Check progress
/sp-validate # Validate work
/sp-continue feature-id # Switch to existing feature
/sp-decompose spec-id # Decompose specifications
/sp-clarify spec-id # Clarify requirements
/sp-llm-enforce [action] # LLM compliance enforcementπ― IMPORTANT: Always start with /sp-pulse - this is your entry point to everything!
SpecPulse uses a revolutionary CLI-First with AI Enhancement approach:
You: /sp-spec "OAuth2 authentication system"
β
Step 1: CLI Creates Foundation (Always Works)
specpulse spec create "OAuth2 authentication system"
β Creates empty spec file
β Adds metadata and structure
β Updates project directories
β
Step 2: AI Enhances Content (Safe Operations)
AI reads created file
AI adds detailed requirements and technical specs
AI enhances with comprehensive documentation
β
Result: Complete, detailed specification ready for development
β CLI First: Creates reliable, cross-platform structure that never fails β AI Enhanced: Leverages AI for what it does best - detailed content creation β Fallback Protection: Work continues even if CLI fails β Platform Independence: Works identically on Windows, macOS, Linux β Safe Operations: AI only works on files CLI has created
Let's build your first feature with SpecPulse step by step:
# Install SpecPulse
pip install specpulse
# Create your first project (choosing your AI platform)
specpulse init my-awesome-project --ai claude
cd my-awesome-projectProject Structure Created:
my-awesome-project/
βββ .specpulse/ # All your project data (specs, plans, tasks)
βββ .claude/ # Claude Code commands (auto-deployed)
βββ README.md
βββ package.json
# THE WAY TO START - AI-Powered Approach
# In your AI assistant (Claude, Gemini, Windsurf, Cursor, etc.):
/sp-pulse user-authenticationWhat happens:
- β
Creates feature directory:
specs/001-user-authentication/ - β Sets up project context and metadata
- β Updates current working feature tracking
- β Prepares structured environment for development
- β Deploys AI platform-specific commands automatically
# In your AI assistant (Claude, Gemini, Windsurf, Cursor, etc.):
/sp-spec "OAuth2 authentication with JWT tokens"AI-powered specification includes:
- π― Problem Statement: Clear business problem and solution overview
- π§ Functional Requirements: Complete feature functionality
- π Security Requirements: Authentication, authorization, data protection
- π¨ User Experience: UI/UX considerations and user flows
- π API Design: Endpoint specifications and data contracts
- πΎ Data Models: Database schemas and relationships
- β Acceptance Criteria: How to verify the implementation works
# In your AI assistant (works on all platforms):
/sp-planAI generates comprehensive plan:
- ποΈ Architecture Decisions: Technology choices and system design
- π File Structure: What files need to be created/modified
- π Implementation Steps: Sequential development approach
- π Dependencies: What needs to be built first
- β±οΈ Time Estimates: Realistic development timelines
- π§ͺ Testing Strategy: How to ensure quality and reliability
# In your AI assistant:
/sp-taskAI creates actionable tasks:
---
id: task-001
status: todo
title: "Set up authentication middleware"
description: |
- What problem does this solve?: Foundation for secure route protection
- Why is this necessary?: All protected endpoints require authentication
- How will this be done?: Express middleware with JWT validation
- When is this complete?: Middleware successfully validates tokens
files_touched:
- path: src/middleware/auth.js
reason: "JWT token validation middleware"
goals: ["Secure authentication middleware", "JWT token validation"]
success_criteria: ["Valid tokens pass", "Invalid tokens are rejected"]
dependencies: []
next_tasks: ["task-002", "task-003"]
---# Execute next task and continue automatically
/sp-execute
# OR execute ALL tasks without stopping
/sp-execute all
# OR execute specific task
/sp-execute task-001What happens during execution:
- π― Task Selection: Finds next pending task
- π Status Update: Marks task as "in-progress"
- π» Implementation: AI writes the actual code
- β Validation: Ensures requirements are met
- π Documentation: Updates task status and moves to next
# Check current progress
/sp-status
# Validate all work completed
/sp-validate# 1. Initialize payment feature
/sp-pulse payment-system
# 2. Create comprehensive specification
/sp-spec "Stripe integration with payment processing, refunds, and subscription management"
# 3. Generate detailed implementation plan
/sp-plan
# 4. Break into development tasks
/sp-task
# 5. Execute all tasks continuously
/sp-execute all
# 6. Review progress
/sp-statusResult in 30 minutes:
- β Complete payment specification (12 pages)
- β Implementation plan with 24 tasks
- β Stripe integration code
- β Payment processing logic
- β Refund system
- β Subscription management
- β Error handling and logging
- β Test suite with 95% coverage
# Team collaboration with different AI platforms:
# Backend developer uses Claude Code:
/sp-pulse realtime-chat
/sp-spec "WebSocket-based chat with rooms, typing indicators, and file sharing"
# Frontend developer uses Gemini CLI:
/sp-continue 001-chat
/sp-task
# DevOps engineer uses Windsurf:
/sp-execute task-005 # WebSocket server setupResult: Seamless collaboration across different AI platforms
Initialize a new feature with intelligent suggestions
# Examples
/sp-pulse user-authentication
/sp-pulse payment-gateway
/sp-pulse real-time-notificationsWhat it does:
- Creates feature directory structure
- Detects project type (web, mobile, API, etc.)
- Suggests specification approaches
- Sets up project context and metadata
Switch to existing feature
# Examples
/sp-continue 001-auth
/sp-continue payment-system
/sp-continue 003-notificationsCreate and expand detailed specification
# Examples
/sp-spec "User authentication with OAuth2 and JWT"
/sp-spec "REST API for product catalog with filtering"
/sp-spec "Real-time chat with WebSocket and rooms"
# With explicit action
/sp-spec create "Payment processing system"
/sp-spec update "Add error handling requirements"
/sp-spec validate
/sp-spec clarifyAI generates complete specification with:
- Business problem and solution overview
- Functional and non-functional requirements
- API contracts and data models
- Security and performance considerations
- User experience and interface design
- Acceptance criteria and success metrics
Validate specification completeness
# Validate current feature's specifications
/sp-spec validateClarify and expand specification details
# Clarify current feature's specifications
/sp-spec clarifyGenerate and expand implementation plan
# Generate plan (default)
/sp-plan
# With explicit action
/sp-plan generate
/sp-plan validate
/sp-plan optimizeAI creates:
- Architecture decisions and technology choices
- File structure and organization
- Implementation steps in logical order
- Dependencies and integration points
- Risk assessment and mitigation strategies
Validate implementation plan
# Validate current feature's plan
/sp-plan validateBreak plan into detailed, actionable tasks
# Break down into tasks (default)
/sp-task
# With explicit action
/sp-task breakdown
/sp-task update
/sp-task status
/sp-task validateEach task includes:
- Clear problem statement and necessity
- Step-by-step implementation approach
- Files to be created/modified with reasons
- Success criteria and acceptance tests
- Dependencies and task relationships
- Risk assessment and mitigation
- MoSCoW priority breakdown (Must, Should, Could, Won't)
Validate task completeness and correctness
# Validate current feature's tasks
/sp-task validateExecute tasks continuously without stopping
# Execute next pending task
/sp-execute
# Execute ALL pending tasks non-stop
/sp-execute all
# Execute specific task
/sp-execute task-001
/sp-execute auth-middlewareExecution behavior:
- Marks task as in-progress
- Implements requirements with code
- Validates against success criteria
- Updates task status to completed
- Automatically moves to next task
- Continues until all tasks done or blocked
Show comprehensive project progress
# Shows:
# - Overall completion percentage
# - Feature-by-feature progress
# - Task status breakdown (pending/in-progress/completed/blocked)
# - Current blockers and issues
# - Time estimates and velocity
# - Quality metrics and test coverageValidate all project artifacts
# Validates:
# - Specification completeness and quality
# - Plan feasibility and coverage
# - Task breakdown and dependencies
# - Code quality and test coverage
# - Documentation and standards complianceDecompose specification into components
# Examples
/sp-decompose spec-001
/sp-decompose auth-specUseful for:
- Microservices architecture design
- Component-based development
- Team distribution and parallel work
- Complex feature breakdown
Clarify requirements and resolve ambiguities
# Examples
/sp-clarify spec-001
/sp-clarify payment-specHelps resolve:
- Unclear requirements
- Missing details
- Conflicting specifications
- Implementation questions
Switch to existing feature and continue work
# Examples
/sp-continue 001-auth
/sp-continue payment-system
/sp-continue 003-notificationsUse this to:
- Switch between different features
- Resume work on existing features
- Change context in multi-feature projects
LLM compliance enforcement and monitoring
# Start enforcement session (default)
/sp-llm-enforce
# Check compliance status
/sp-llm-enforce status
# Validate operations against rules
/sp-llm-enforce validate
# End enforcement session
/sp-llm-enforce endEnsures:
- Directory traversal protection
- File operation security
- Content validation and sanitization
- Compliance with SpecPulse rules
- Audit trail generation
your-project/
βββ .specpulse/ # All SpecPulse data (git-tracked)
β βββ specs/ # Feature specifications
β β βββ 001-user-auth/
β β β βββ spec-001.md # Main specification
β β β βββ decomposition/ # Service breakdown (if microservices)
β β β βββ references/ # External docs and links
β β βββ 002-payment-system/
β β βββ ...
β βββ plans/ # Implementation plans
β β βββ 001-user-auth/
β β β βββ plan-001.md # Implementation plan
β β β βββ alternatives.md # Alternative approaches
β β βββ ...
β βββ tasks/ # Development tasks
β β βββ 001-user-auth/
β β β βββ task-001.md # Individual tasks
β β β βββ task-002.md
β β β βββ task-summary.md # Progress tracking
β β βββ ...
β βββ memory/ # Project context and decisions
β β βββ context.md # Current feature context
β β βββ decisions.md # Architecture decisions
β β βββ learning.md # Team knowledge base
β βββ templates/ # Customizable templates
β βββ spec.md # Specification template
β βββ plan.md # Planning template
β βββ task.md # Task template
βββ .claude/ # Claude Code commands
βββ .gemini/ # Gemini CLI commands
βββ .windsurf/ # Windsurf AI workflows
βββ .cursor/ # Cursor AI commands
βββ .github/prompts/ # GitHub Copilot prompts
βββ .opencode/command/ # OpenCode AI commands
βββ .crush/commands/sp/ # Crush AI commands
βββ .qwen/commands/ # Qwen Code commands
βββ your-project-files/ # Your actual project code
β Tracked in Git (Important):
- All specifications (
.specpulse/specs/) - Implementation plans (
.specpulse/plans/) - Task definitions and progress (
.specpulse/tasks/) - Project memory and decisions (
.specpulse/memory/) - Custom templates (
.specpulse/templates/)
π Generated/Managed (Don't edit manually):
- AI command deployments (
.claude/,.gemini/, etc.) - Temporary working files
- Auto-generated documentation
# Install SpecPulse
pip install specpulse
# Create new project with AI integration
specpulse init my-project --ai claude
# Add AI support to existing project
specpulse init --here --ai gemini
# Check system health
specpulse doctorβ DO:
- Start with clear problem statement
- Include specific acceptance criteria
- Define user stories and use cases
- Specify error handling and edge cases
- Include performance and security requirements
- Add API contracts and data models
- Consider accessibility and internationalization
β DON'T:
- Write vague requirements like "make it fast"
- Skip error handling scenarios
- Forget about testing and validation
- Ignore user experience considerations
- Leave security as an afterthought
β DO:
- Break tasks into small, manageable chunks (2-8 hours each)
- Include clear success criteria for each task
- Define dependencies between tasks
- Consider parallel work opportunities
- Include testing and documentation tasks
- Plan for integration and deployment
β DON'T:
- Create tasks that are too large (over 16 hours)
- Forget about setup and configuration tasks
- Skip validation and testing tasks
- Ignore deployment and monitoring
- Leave dependencies undefined
β DO:
- Use feature branches for each feature
- Commit specifications and plans
- Review specifications as a team
- Use consistent task numbering
- Track progress with
/sp-status - Hold regular planning meetings
β DON'T:
- Work without specifications
- Skip code reviews
- Ignore task dependencies
- Forget about documentation
- Work in isolation without communication
# 1. Create monolithic specification first
/sp-pulse user-management
/sp-spec "Complete user management system with profiles, authentication, and preferences"
# 2. Decompose into services
/sp-decompose spec-001
# This creates:
# - User Service (profiles, preferences)
# - Auth Service (authentication, authorization)
# - Notification Service (emails, push notifications)
# - API Gateway (routing, rate limiting)
# 3. Generate service-specific tasks
/sp-task
# 4. Execute by service team
/sp-execute task-001 # User Service setup
/sp-execute task-005 # Auth Service setup# 1. Document existing system
/sp-pulse legacy-modernization
/sp-spec "Modernize legacy monolith to microservices: current state analysis and migration strategy"
# 2. Create migration plan
/sp-plan
# 3. Break into migration phases
/sp-task
# 4. Execute migration incrementally
/sp-execute task-001 # Phase 1: Strangler Fig pattern setup
/sp-execute task-002 # Phase 2: Extract user management# 1. Design API first
/sp-pulse product-catalog-api
/sp-spec "REST API for product catalog with CRUD operations, search, filtering, and pagination"
# 2. Generate OpenAPI specification
# (AI includes OpenAPI spec in the generated specification)
# 3. Create implementation plan
/sp-plan
# 4. Implement API
/sp-execute all# 1. Define mobile app features
/sp-pulse mobile-expense-tracker
/sp-spec "Mobile expense tracking app with receipt scanning, categorization, and reporting"
# 2. Plan platform-specific implementation
/sp-plan
# 3. Break into platform-specific tasks
/sp-task
# 4. Develop by platform
/sp-execute task-001 # React Native setup
/sp-execute task-005 # iOS-specific features
/sp-execute task-010 # Android-specific featuresCreate your own templates for consistent project standards:
# Create custom specification template
specpulse template create enterprise-spec --type spec
# Edit the template
# .specpulse/templates/enterprise-spec.md
# Use custom template
/sp-spec "New feature" --template enterprise-specCustomize SpecPulse behavior for your project:
# .specpulse/config.yml
project:
name: "My Enterprise App"
type: "web"
team_size: 5
specifications:
template: "enterprise-spec"
required_sections:
- "problem_statement"
- "requirements"
- "api_design"
- "security"
- "testing"
tasks:
default_estimates:
simple: 4 # hours
medium: 8 # hours
complex: 16 # hours
ai_preferences:
code_style: "enterprise"
testing_framework: "jest"
documentation_style: "jsdoc"Challenge: Build and launch MVP in 6 weeks with 2 developers Solution: SpecPulse with Claude Code integration
Week 1: Feature Planning
/sp-pulse user-management
/sp-pulse product-catalog
/sp-pulse order-processing
Week 2-3: Specification and Planning
/sp-spec "Complete user management system"
/sp-spec "Product catalog with search and filtering"
/sp-spec "Order processing with payment integration"
/sp-plan
Week 4-5: Development
/sp-task
/sp-execute all
Week 6: Testing and Deployment
/sp-validateResult: MVP launched on time with all core features, 95% test coverage
Challenge: Migrate 500k+ LOC monolith to microservices over 12 months Solution: SpecPulse with team-wide AI integration
Phase 1: Analysis (1 month)
/sp-pulse legacy-analysis
/sp-spec "Complete monolith analysis and service decomposition strategy"
Phase 2: Pilot (2 months)
/sp-pulse user-service-migration
/sp-execute all
Phase 3: Full Migration (9 months)
# Multiple teams working in parallel
# Each team uses /sp-continue to work on their servicesResult: Migration completed on schedule with zero downtime, improved performance by 300%
Challenge: Coordinate 50+ contributors across different time zones Solution: SpecPulse with GitHub Copilot integration
# Contributors use GitHub Copilot with /sp-* commands
# All specifications and plans tracked in .specpulse/
# Clear task breakdown enables parallel work
/sp-status shows overall project progressResult: Development velocity increased by 200%, contributor satisfaction improved
- Python: 3.11 or higher
- Operating System: Windows 10+, macOS 10.15+, Ubuntu 18.04+
- Memory: 4GB RAM minimum, 8GB recommended
- Storage: 100MB for SpecPulse + project data
Compatible AI Platforms:
- Claude Code (Anthropic)
- Gemini CLI (Google)
- Windsurf AI (Windsurf)
- Cursor AI (Cursor)
- GitHub Copilot (Microsoft)
- OpenCode AI (OpenCode)
- Crush AI (Crush)
- Qwen Code (Alibaba)
Project Initialization:
- Small project (< 100 files): < 2 seconds
- Medium project (100-1000 files): < 5 seconds
- Large project (1000+ files): < 10 seconds
Specification Generation:
- Simple feature: 30-60 seconds
- Complex feature: 2-5 minutes
- Enterprise system: 5-15 minutes
Task Breakdown:
- Small plan (< 10 tasks): < 30 seconds
- Medium plan (10-50 tasks): 1-2 minutes
- Large plan (50+ tasks): 2-5 minutes
Recommended Project Sizes:
- Small Projects: < 50 tasks, < 10 specifications
- Medium Projects: 50-500 tasks, 10-50 specifications
- Large Projects: 500+ tasks, 50+ specifications
- Enterprise: Multiple teams, 1000+ tasks
Performance Optimization:
- Use SSD storage for faster file operations
- Enable file system indexing for large projects
- Use parallel task execution for independent tasks
- Regular cleanup of old completed tasks
SpecPulse respects your privacy:
- β No external API calls - Everything runs locally
- β No data transmission - Your specifications never leave your system
- β No telemetry - We don't track usage or project data
- β Local AI integration - AI assistants work with your local files only
File Security:
- All files created with appropriate permissions
- Sensitive data handling in specifications
- Secure temporary file management
- Path traversal protection
Code Security:
- Input validation and sanitization
- Template injection protection
- Safe file operation practices
- Dependency security scanning
Team Security:
- Git-based access control
- Specification review workflows
- Audit trail for all changes
- Role-based access patterns
For Enterprise Deployments:
- Air-gapped installation support
- Custom template encryption
- Audit logging and compliance
- Integration with enterprise SSO
- Custom security policies
Issue: Command not found
# Solution: Verify installation
pip show specpulse
python -m specpulse --version
# Reinstall if needed
pip uninstall specpulse
pip install specpulseIssue: AI commands not working
# Solution: Check AI platform integration
specpulse doctor --fix
# Manually deploy commands
specpulse init --here --ai claude # Redeploy commandsIssue: File permission errors
# Solution: Check permissions
ls -la .specpulse/
chmod 755 .specpulse/
chmod 644 .specpulse/specs/*.mdIssue: Git integration problems
# Solution: Initialize git repository
git init
git add .
git commit -m "Initial SpecPulse setup"Self-Service:
# General help
specpulse --help
# Command-specific help
specpulse feature --help
specpulse spec --help
# Health check and diagnostics
specpulse doctorCommunity Support:
- GitHub Issues: Report bugs and request features
- GitHub Discussions: Community questions and discussions
- Documentation: Available in the project repository
Enable detailed logging:
# Enable debug mode
export SPECPULSE_DEBUG=1
# Or for Windows
set SPECPULSE_DEBUG=1
# Run commands with debug output
specpulse doctorDebug file locations:
- Logs:
.specpulse/logs/debug.log - Cache:
.specpulse/cache/ - Temp:
.specpulse/tmp/
# Clone repository
git clone https://github.com/specpulse/specpulse.git
cd specpulse
# Create virtual environment
python -m venv venv
source venv/bin/activate # On Windows: venv\Scripts\activate
# Install development dependencies
pip install -e ".[dev]"
# Run tests
pytest tests/
# Run with coverage
pytest --cov=specpulse --cov-report=htmlWe welcome contributions!
Areas for contribution:
- π Bug fixes: Help us squash bugs
- β¨ New features: Propose and implement new functionality
- π Documentation: Improve documentation and examples
- π§ͺ Testing: Add test cases and improve test coverage
- π§ Tools: Development tools and integrations
Development Process:
- Fork the repository
- Create feature branch (
git checkout -b feature/amazing-feature) - Make your changes
- Add tests for new functionality
- Ensure all tests pass (
pytest) - Commit your changes (
git commit -m 'Add amazing feature') - Push to branch (
git push origin feature/amazing-feature) - Open Pull Request
Python Code Style:
- Follow PEP 8 style guidelines
- Use Black for code formatting
- Use isort for import sorting
- Type hints required for new code
- Maximum line length: 88 characters
Documentation Style:
- Use clear, concise language
- Include code examples
- Follow Markdown formatting standards
- Include proper section headers
- Add ALT text for images
Testing Standards:
- Write unit tests for all new functionality
- Aim for high code coverage (>90%)
- Use descriptive test names
- Include edge case testing
- Mock external dependencies
SpecPulse is released under the MIT License. You can find the full license text in the LICENSE file.
What this means:
- β Commercial use allowed
- β Modification allowed
- β Distribution allowed
- β Private use allowed
- β Liability and warranty disclaimed
# 1. Install SpecPulse
pip install specpulse
# 2. Create your first project
specpulse init my-project --ai claude
cd my-project
# 3. Start your first feature (THE ENTRY POINT)
/sp-pulse amazing-feature
# 4. Create your first specification
/sp-spec "Build an amazing feature that will wow users"
# 5. Plan and execute
/sp-plan
/sp-task
/sp-execute all
# 6. Celebrate your success!
π- GitHub: github.com/specpulse/specpulse
- Developer: Ersin KOΓ β’ x.com/ersinkoc
- Website: specpulse.xyz for full documentation
- Support: Create an issue
- Contact: info@specpulse.xyz for business inquiries
**β Star us on GitHub β’ π¦ Install from PyPI **
Made with β€οΈ by developers who believe in specifications-first development
SpecPulse v2.7.5 - Building the future of software development, together.