AutoDocs MCP Server

# Agent Context Architecture **Purpose**: Define how AI agents access and share context across the multi-project ecosystem **Scope**: Context management for AI development agents working across projects **Updated**: August 11, 2025 ## 🤖 Agent Context Management Philosophy ### Core Principles - **Project Autonomy**: Each project maintains its own specific context in CLAUDE.md - **Shared Foundation**: Common knowledge and standards accessible to all agents - **Context Clarity**: Clear boundaries between project-specific and shared context - **Documentation as Code**: All context is version-controlled and deployable ### Context Hierarchy ``` Shared Context (All Agents) ├── Development Standards & Conventions ├── Technology Stack Decisions ├── Quality & Testing Standards ├── Architecture Patterns └── Integration Protocols Project-Specific Context (Per-Project Agents) ├── Project Goals & Scope ├── Technical Implementation Details ├── Project-Specific Standards ├── Current Status & Priorities └── Historical Context & Decisions ``` --- ## 📁 Shared Context Repository ### What All Agents Need to Know #### Development Standards (Cross-Project) - **Code Quality**: Ruff formatting, MyPy type checking, pre-commit hooks - **Testing Patterns**: pytest ecosystem, >90% coverage, pytest-mock patterns - **Documentation**: Markdown standards, API documentation requirements - **Version Control**: Conventional commits, GitFlow branching, semantic versioning #### Technology Stack (Cross-Project) - **Python Ecosystem**: Python 3.12+, uv package management, FastAPI/FastMCP - **Quality Tools**: Ruff, MyPy, pytest ecosystem, pre-commit - **Deployment**: PyPI packages, GitHub Actions CI/CD - **Monitoring**: Health checks, structured logging, metrics collection #### Architecture Patterns (Cross-Project) - **MCP Protocol**: Standard for AI tool integration - **Async Design**: Full async/await patterns for scalability - **Error Handling**: Graceful degradation, structured error responses - **Security**: Input validation, secure error handling, dependency scanning --- ## 🎯 Project-Specific Context Boundaries ### AutoDocs MCP Server Context **Unique Knowledge**: - PyPI API integration patterns - Documentation caching strategies - MCP tool implementation specifics - Production monitoring and reliability patterns **Shared with Other Projects**: - MCP protocol expertise - Python async patterns - Testing and quality standards ### Documentation Site Context **Unique Knowledge**: - Content strategy and information architecture - Static site generation and deployment - User experience and content optimization **Shared with Other Projects**: - Web development standards - Documentation writing principles - Deployment and CI/CD patterns ### Task Graph System Context **Unique Knowledge**: - Multi-agent coordination patterns - AI agent communication protocols - Research methodology and experimentation **Shared with Other Projects**: - AI integration patterns - Research and prototyping approaches - Architecture design principles --- ## 🔄 Context Sharing Mechanisms ### CLAUDE.md Integration Each project's `CLAUDE.md` file contains: - **Project-Specific Instructions**: Unique to that project's needs - **Shared Context References**: Links to cross-project standards and patterns - **Integration Points**: How this project connects with others ### Cross-Project Knowledge Base Located in `planning/cross-project/`: - **Standards Documentation**: Accessible to all project agents - **Technology Decisions**: Shared technology stack and rationale - **Architecture Patterns**: Reusable design patterns and solutions - **Best Practices**: Lessons learned and proven approaches ### Documentation Site Context The planning folder serves as source material for: - **Public Documentation**: Human-readable project descriptions and architecture - **Agent Training Data**: Examples and patterns for future AI agent training - **Community Resource**: Open source community reference and contribution guide --- ## 📊 Context Management Strategies ### Avoiding Context Duplication - **Single Source of Truth**: Each piece of knowledge has one authoritative location - **Reference Pattern**: Projects reference shared context rather than duplicating - **Consistent Updates**: Changes to shared context automatically benefit all projects ### Context Evolution - **Version Control**: All context changes tracked in Git - **Change Impact**: Consider how context changes affect all projects - **Migration Strategy**: Plan for updating project-specific context when shared context evolves ### Agent Onboarding When agents work on new projects: 1. **Read Shared Context**: Understand cross-project standards and patterns 2. **Read Project Context**: Understand specific project goals and implementation 3. **Identify Integration Points**: Understand how this project connects with others 4. **Apply Consistent Patterns**: Use shared patterns while respecting project uniqueness --- ## 🏗️ Implementation Architecture ### Directory Structure for Agents ``` planning/ ├── cross-project/ # Shared agent context │ ├── development_standards.md # How to write code across all projects │ ├── technology_stack.md # Shared technology decisions and rationale │ ├── architecture_patterns.md # Reusable design patterns │ └── integration_protocols.md # How projects connect and share functionality ├── projects/ # Project-specific agent context │ ├── autodocs-mcp/CLAUDE.md # Project-specific agent instructions │ ├── documentation-site/CLAUDE.md # Project-specific agent instructions │ └── task-graph-system/CLAUDE.md # Project-specific agent instructions └── PUBLIC_DOCUMENTATION/ # Human-readable docs site content ``` ### Context Access Pattern 1. **Agent starts on project** → reads project's CLAUDE.md 2. **CLAUDE.md references** → cross-project standards and patterns 3. **Agent applies** → consistent patterns with project-specific adaptation 4. **Agent contributes** → improvements to shared context when applicable --- ## 🌐 Documentation Site Architecture ### Site Content Strategy The planning folder content is structured for deployment as a documentation site: **Target Audience**: - Developers interested in the projects - AI researchers studying multi-agent systems - Open source contributors - Future AI agents learning from examples **Content Organization**: - **Project Overviews**: High-level descriptions and current status - **Architecture Documentation**: Technical system designs and patterns - **Development Philosophy**: Principles and approaches that guide development - **Standards and Conventions**: How code is written and projects are managed ### Site Structure ``` docs/ ├── index.md # Portfolio overview and navigation ├── projects/ # Individual project documentation │ ├── autodocs-mcp/ # AutoDocs project docs │ ├── documentation-site/ # Documentation project docs │ └── task-graph-system/ # Research project docs ├── architecture/ # Cross-project technical architecture ├── development/ # Standards, practices, and conventions └── philosophy/ # Principles and approaches ``` --- ## 🔮 Future Evolution ### Scaling Context Management As the project portfolio grows: - **Context Categorization**: More granular organization of shared vs specific context - **Agent Specialization**: Different types of agents with different context needs - **Automated Context**: Tools to automatically maintain context consistency - **Community Context**: External contributions to shared knowledge base ### Advanced Agent Coordination Future possibilities: - **Context APIs**: Programmatic access to context for agent coordination - **Dynamic Context**: Context that adapts based on current project states - **Cross-Agent Learning**: Agents learning from each other's experiences - **Intelligent Context Routing**: Automatically providing relevant context to agents --- *Agent context architecture established: August 11, 2025* *Context management approach: Multi-project AI agent coordination* *Documentation site deployment: Planned for GitHub Pages*