Spec Workflow MCP

# Technical Documentation > **Quick Reference**: Jump to what you need most → [Tools API](api-reference.md) | [Architecture](architecture.md) | [Developer Guide](developer-guide.md) | [Troubleshooting](troubleshooting.md) ## 📋 Table of Contents ### Core Documentation - **[Architecture Overview](architecture.md)** - System design, components, and data flow - **[MCP Tools API Reference](api-reference.md)** - Complete tool documentation with examples - **[Developer Workflow Guide](developer-guide.md)** - Step-by-step development workflows - **[Context Management](context-management.md)** - How context switching and caching works - **[File Structure](file-structure.md)** - Project organization and directory layout - **[Dashboard System](dashboard.md)** - Web dashboard and real-time features - **[Troubleshooting & FAQ](troubleshooting.md)** - Common issues and solutions ### Quick Start Guides - **[Setting Up Development Environment](setup.md)** - Get up and running quickly - **[Contributing Guidelines](contributing.md)** - How to contribute to the project - **[Testing Guide](testing.md)** - Running tests and writing new ones ## 🚀 Quick Start ### For AI Assistant Integration ```json { "mcpServers": { "spec-workflow": { "command": "npx", "args": ["-y", "@pimzino/spec-workflow-mcp@latest", "/path/to/project", "--AutoStartDashboard"] } } } ``` ### For Local Development ```bash # Clone and setup git clone <repository-url> cd spec-workflow-mcp npm install # Start development server npm run dev # Build for production npm run build ``` ## 🔍 Comprehensive Capability Analysis ### Critical Questions Answered Based on comprehensive codebase analysis, here are definitive answers to key technical questions: #### **Question 1: Web Scraping & Research Capabilities** **Answer: No independent web scraping - leverages LLM's built-in web search** | Aspect | This MCP | Other AI Agents | Expansion Opportunity | |--------|----------|----------------|---------------------| | **Web Scraping** | ❌ No independent capability | ✅ Custom scrapers (Puppeteer, Playwright) | 🔮 Could add structured scraping tools | | **API Research** | ❌ Relies on LLM's web search | ✅ Direct API integrations | 🔮 Could add GitHub, Stack Overflow APIs | | **Research Caching** | ❌ No research persistence | ✅ Advanced caching systems | 🔮 Could cache LLM research results | | **Data Sources** | ✅ LLM's vast training data + real-time web | ❌ Limited to configured sources | ✅ Best of both worlds | #### **Question 2: AI Calls & Context Window Management** **Answer: Pure MCP - uses only connected LLM, no independent AI calls** | Aspect | This MCP | Other AI Agents | Expansion Opportunity | |--------|----------|----------------|---------------------| | **AI Service Calls** | ❌ No independent AI calls | ✅ Multiple AI model integration | 🔮 Could add specialized AI services | | **Context Management** | ❌ No LLM context manipulation | ✅ Advanced context strategies | 🔮 Could add context optimization | | **Memory Management** | ❌ File-based only | ✅ Vector databases, embeddings | 🔮 Could add persistent memory | | **Multi-Model Usage** | ❌ Single LLM connection | ✅ GPT-4 + Claude + Gemini | 🔮 Could add model routing | #### **Question 3: Document Planning Process** **Answer: Template-guided LLM intelligence - no separate AI planning** | Aspect | This MCP | Other AI Agents | Expansion Opportunity | |--------|----------|----------------|---------------------| | **Planning Intelligence** | ✅ LLM reasoning with templates | ✅ Dedicated planning AI | 🔮 Could add adaptive workflows | | **Template System** | ✅ Static but comprehensive | ❌ Often no structured templates | ✅ Structured advantage | | **Workflow Adaptation** | ❌ Fixed sequence | ✅ Dynamic workflow generation | 🔮 Could add LLM-powered workflows | | **Project Analysis** | ✅ LLM analyzes project context | ✅ Specialized analysis tools | 🔮 Could add deep code analysis | #### **Question 4: Auto Review Process** **Answer: Human-only approval system - no automated AI review** | Aspect | This MCP | Other AI Agents | Expansion Opportunity | |--------|----------|----------------|---------------------| | **Review Automation** | ❌ Human approval required | ✅ Multi-stage AI review | 🔮 Could add optional AI gates | | **Quality Assurance** | ✅ LLM quality + Human oversight | ❌ AI-only (potential errors) | ✅ Best quality control | | **Approval Workflows** | ✅ Dashboard/VS Code integration | ❌ Often CLI-only | ✅ Superior UX | | **Review Intelligence** | ✅ LLM can suggest improvements | ✅ Specialized review models | 🔮 Could add review templates | #### **Question 5: Best Practice Standards** **Answer: LLM built-in knowledge - no external standards fetching** | Aspect | This MCP | Other AI Agents | Expansion Opportunity | |--------|----------|----------------|---------------------| | **Standards Source** | ✅ LLM's vast training knowledge | ✅ External standards APIs | 🔮 Could add standards integration | | **Currency** | ✅ LLM can web search for latest | ❌ Static configurations | ✅ Always current | | **Customization** | ❌ No project-specific standards | ✅ Custom rule engines | 🔮 Could add org standards | | **Best Practices** | ✅ Industry-wide via LLM | ❌ Limited to pre-configured | ✅ Comprehensive coverage | ### Competitive Positioning Analysis **Strengths vs Other AI Agents:** ```typescript interface CompetitiveAdvantages { humanOversight: "Mandatory approval prevents runaway AI behavior"; llmLeverage: "Uses full power of connected LLM without limitations"; structuredOutput: "Templates ensure consistent, professional documentation"; realTimeUI: "Dashboard and VS Code integration for seamless workflow"; simplicity: "No complex setup or API key management required"; reliability: "Proven workflow sequence with validation and error handling"; } ``` **Current Limitations vs Market Leaders:** ```typescript interface LimitationsAnalysis { automationLevel: "Less automated than fully autonomous agents"; integrationEcosystem: "Limited external service integrations"; multiProject: "Single project scope vs enterprise-wide solutions"; aiDiversity: "Single LLM vs multi-model approaches"; workflowFlexibility: "Fixed sequence vs adaptive workflows"; } ``` **Expansion Opportunities Identified:** ```typescript interface ExpansionRoadmap { immediateWins: { githubIntegration: "PR creation, issue sync, code analysis"; qualityGates: "Optional automated quality checks"; templateDynamism: "Project-type aware template selection"; }; mediumTerm: { multiProjectSupport: "Enterprise dashboard for multiple projects"; advancedIntegrations: "Jira, Confluence, Slack notifications"; workflowCustomization: "Configurable workflow sequences"; }; longTerm: { aiOrchestration: "Multi-agent coordination capabilities"; predictiveAnalytics: "Project success prediction and risk analysis"; enterpriseFeatures: "SSO, compliance, audit trails"; }; } ``` ## ⚠️ Technical Limitations & Capabilities ### What This MCP Does NOT Do **No Independent External Calls**: - ❌ No separate web scraping or API calls by the MCP server - ❌ No independent external research by the MCP server - ❌ No direct calls to AI services from the MCP server - ✅ Leverages connected LLM's built-in web search and knowledge **No Separate AI Service Integration**: - ❌ No additional calls to OpenAI, Anthropic, or other AI services - ❌ No independent AI processing outside the connected LLM - ❌ No separate AI models or services - ✅ Uses only the LLM provided through MCP connection **No Context Window Management**: - ❌ Does not extend or manage AI client context windows - ❌ No conversation history or memory management - ❌ No cross-session AI context preservation - ✅ Provides structured project data for AI client consumption **Human-Only Approval System**: - ❌ No automated AI-powered document review - ❌ No AI-based approval recommendations - ❌ Verbal approval not accepted - ✅ All approvals require dashboard or VS Code interaction ### What This MCP Excels At **Leveraging LLM Built-in Capabilities**: - ✅ Provides structured templates for LLM to fill with intelligent content - ✅ Supplies project context for LLM analysis and understanding - ✅ Enables LLM to use its built-in knowledge for best practices - ✅ Allows LLM to perform web research when generating content **Structured Workflow Enforcement**: - ✅ Enforces spec-driven development sequence - ✅ Template-based document structure for consistent LLM output - ✅ Workflow validation and blocking - ✅ Human oversight integration for LLM-generated content **Intelligent Project Data Management**: - ✅ Efficient context loading for LLM consumption - ✅ Real-time file watching and updates - ✅ Cross-platform path handling - ✅ Structured project organization that LLM can understand **Enhanced Developer Experience**: - ✅ Web dashboard for reviewing LLM-generated content - ✅ VS Code extension integration - ✅ Real-time WebSocket updates - ✅ Comprehensive error handling ## 🎯 Key Concepts ### MCP Tools The server provides 12 MCP tools for spec-driven development: - **Workflow Tools**: `spec-workflow-guide`, `steering-guide` - **Content Tools**: `create-spec-doc`, `create-steering-doc`, `get-template-context` - **Search Tools**: `get-spec-context`, `get-steering-context`, `spec-list` - **Status Tools**: `spec-status`, `manage-tasks` - **Approval Tools**: `request-approval`, `get-approval-status`, `delete-approval` ### File Organization ``` .spec-workflow/ ├── specs/ # Specification documents ├── steering/ # Project guidance documents ├── approvals/ # Approval workflow data └── session.json # Active session tracking ``` ### Workflow Phases 1. **Requirements** → 2. **Design** → 3. **Tasks** → 4. **Implementation** Each phase requires approval before proceeding to the next. ## 🔧 Development Workflow ### Adding a New MCP Tool 1. Create tool file in `src/tools/` 2. Export tool definition and handler 3. Register in `src/tools/index.ts` 4. Update API documentation 5. Add tests ### Dashboard Development ```bash # Start dashboard in development mode npm run dev:dashboard # Build dashboard assets npm run build:dashboard ``` ### VSCode Extension Development ```bash cd vscode-extension npm install npm run compile # Press F5 in VSCode to launch extension host ``` ## 📚 Documentation Standards - **Code Examples**: Always include working examples - **Error Handling**: Document expected error conditions - **Performance**: Note any performance considerations - **Security**: Highlight security implications - **Breaking Changes**: Mark breaking changes clearly ## 🤝 Getting Help 1. **Check the [Troubleshooting Guide](troubleshooting.md)** first 2. **Search existing [GitHub Issues](https://github.com/Pimzino/spec-workflow-mcp/issues)** 3. **Create a new issue** with detailed reproduction steps 4. **Join the community** for real-time support --- ## 📊 Technical Architecture Summary ### Pure MCP Server Design This project implements a **pure Model Context Protocol (MCP) server** that: | Aspect | Implementation | Details | |--------|---------------|----------| | **AI Integration** | Pure MCP server | Leverages connected LLM's built-in capabilities | | **Web Research** | LLM built-in capability | LLM performs web search using its built-in features | | **Context Management** | File-based structure | No LLM context window management | | **Content Generation** | LLM-powered with templates | LLM fills templates using built-in knowledge & search | | **Planning Process** | LLM reasoning + workflow validation | LLM plans content, MCP enforces structure | | **Review System** | Human approval only | Dashboard/VS Code integration for LLM output | | **Best Practices** | LLM built-in knowledge | LLM applies best practices from its training | | **External Calls** | NPM version check only | All other capabilities through connected LLM | ### Key Files & Implementation - **MCP Tools**: `src/tools/*.ts` - 13 tools for workflow management - **Templates**: `src/markdown/templates/*.md` - Static document structures - **Approval System**: `src/dashboard/approval-storage.ts` - Human-only review - **Context Loading**: `src/core/*.ts` - File-based context structuring - **Web Dashboard**: `src/dashboard_frontend/` - React-based approval UI ### Performance Characteristics - **Memory Usage**: 50KB templates + 10-100KB per spec context - **File System**: Local `.spec-workflow/` directory only - **Network**: Localhost dashboard + NPM version check - **Scaling**: Linear per project, 50-100 specs recommended - **Security**: Local-only, no external data transmission ## 📊 Market Analysis & Strategic Insights ### Competitive Landscape Analysis **Category 1: Autonomous AI Agents (e.g., AutoGPT, LangChain Agents)** ```typescript interface AutonomousAgents { capabilities: { webScraping: "Advanced - Custom scrapers, API integrations"; aiCalls: "Multiple models, specialized AI services"; automation: "Fully autonomous operation"; integrations: "Extensive third-party ecosystem"; }; limitations: { humanOversight: "Limited or optional"; reliability: "Can go off-track or produce errors"; complexity: "Complex setup, API management"; cost: "High due to multiple AI calls"; }; differentiator: "Full automation vs structured human-guided workflow"; } ``` **Category 2: Development Workflow Tools (e.g., GitHub Copilot, Cursor)** ```typescript interface DevelopmentTools { capabilities: { codeGeneration: "Excellent within editors"; contextAwareness: "Good for code context"; realTimeAssistance: "Integrated development support"; aiPowered: "Built-in LLM capabilities"; }; limitations: { workflowStructure: "Limited structured spec processes"; documentationFocus: "Code-centric, not spec-driven"; approvalProcess: "No formal review workflows"; projectPlanning: "Limited high-level planning"; }; differentiator: "Code-first vs spec-driven development approach"; } ``` **Category 3: Project Management + AI (e.g., Notion AI, Linear)** ```typescript interface ProjectManagementAI { capabilities: { projectTracking: "Excellent project organization"; collaboration: "Team coordination features"; aiAssistance: "AI-powered content generation"; integration: "Extensive third-party connections"; }; limitations: { technicalDepth: "Limited technical specification focus"; workflowEnforcement: "Flexible but not enforced"; developerWorkflow: "Not developer-workflow optimized"; codeIntegration: "Limited code context understanding"; }; differentiator: "General project management vs developer-specific workflows"; } ``` ### Strategic Market Position **Spec-Workflow-MCP's Unique Position:** ```typescript interface MarketPosition { blueOcean: { category: "LLM-Enhanced Structured Development Workflows"; uniqueValue: "Human-supervised LLM intelligence with enforced spec-driven process"; targetUser: "Development teams needing structured processes with AI assistance"; }; competitiveAdvantages: { llmLeverage: "Full LLM power without additional API costs"; humanOversight: "Prevents AI errors through mandatory approval"; structuredProcess: "Enforces proven development methodology"; simplicity: "No complex setup or API key management"; realTimeUI: "Superior user experience with dashboard"; }; marketOpportunities: { enterpriseAdoption: "Companies wanting AI benefits with human control"; consultingFirms: "Standardized processes across client projects"; startups: "Structured development without overhead"; education: "Teaching proper development workflows"; }; } ``` ### Expansion Strategy Insights **Phase 1: Leverage Core Strengths** ```typescript interface Phase1Strategy { buildOnStrengths: { enhanceHumanOversight: "Advanced approval workflows, review templates"; improveStructure: "Dynamic templates, adaptive workflows"; expandLLMUsage: "Better context utilization, smarter suggestions"; }; addressGaps: { basicIntegrations: "GitHub, GitLab, Bitbucket connections"; qualityGates: "Optional automated checks before human review"; teamFeatures: "Multi-developer coordination"; }; } ``` **Phase 2: Strategic Differentiation** ```typescript interface Phase2Strategy { uniqueCapabilities: { hybridIntelligence: "Best of LLM automation + human oversight"; contextMastery: "Superior project context understanding"; processExcellence: "Industry-leading structured workflows"; }; competitiveFeatures: { multiModelSupport: "Support multiple LLM providers"; enterpriseFeatures: "SSO, compliance, audit trails"; aiOrchestration: "Multi-agent coordination while maintaining oversight"; }; } ``` ### Strategic Recommendations for Creators **Immediate Opportunities (0-6 months):** 1. **GitHub Integration**: Leverage LLM to create PRs, analyze codebases 2. **Quality Templates**: Add project-type detection for smarter templates 3. **Team Coordination**: Multi-developer approval workflows 4. **Performance Analytics**: Track spec-to-delivery success rates **Medium-term Differentiators (6-18 months):** 1. **Hybrid AI Workflows**: Optional automated gates with human oversight 2. **Enterprise Dashboard**: Multi-project management interface 3. **Advanced Integrations**: Jira, Slack, Confluence, CI/CD pipelines 4. **Predictive Analytics**: Project risk analysis using LLM insights **Long-term Vision (18+ months):** 1. **AI Orchestration Platform**: Multi-agent coordination with human oversight 2. **Industry Templates**: Specialized workflows for different domains 3. **Compliance Integration**: SOX, GDPR, HIPAA workflow templates 4. **Educational Platform**: Teaching structured development at scale ### Market Validation Insights **This analysis reveals that Spec-Workflow-MCP occupies a unique market position:** - ✅ **Underserved Market**: Structured development workflows with AI enhancement - ✅ **Clear Differentiation**: Human oversight + LLM power combination - ✅ **Expansion Potential**: Multiple clear paths for feature enhancement - ✅ **Strategic Moat**: Proven workflow methodology that competitors would struggle to replicate **Last Updated**: December 2024 | **Version**: 0.0.23