Documentation Generator MCP Server

{ "feature_name": "agent-personas", "description": "Create a team of specialized AI agents that work together to handle complex software development tasks. Each agent has domain expertise and follows a standardized communication protocol to coordinate work, maintain context, and deliver precise results within the MCP ecosystem.", "goal": "Build an MCP server that provides highly-trained specialist agents capable of autonomous task execution with improved efficiency, parallel processing, and expert-level precision across multiple domains (UI/UX, frameworks, backend, testing, documentation, etc.)", "requirements": [ "Two meta-level coordinator agents: Project Leader and Project Assistant", "Project Leader: Breaks down complex requests, delegates to specialists, monitors progress, coordinates between agents, makes architectural decisions", "Project Assistant: Works directly with user as assistant - answers questions, tracks tasks/progress, runs routine checks, gathers info/research, takes notes and documents decisions", "Specialist agents covering multiple domains: Code-focused (debugging, refactoring, testing), Documentation, Planning, Analysis (code review, security, performance), Integration (API, deployment, CI/CD), UI/UX, Framework-specific, and custom domain agents", "Agent training via combination of static configuration (pre-configured prompts) + dynamic context loading (project-specific standards/patterns from docs-mcp)", "Integration as MCP tools (e.g., mcp__agents__invoke_ui_expert)", "Integration as slash commands (e.g., /ui-review, /framework-migrate)", "Support for automatic agent routing (system analyzes task and routes to best agent)", "Each agent has access to: project-specific context (codebase analysis, standards, patterns), framework docs/best practices, anti-patterns to avoid, real-world examples, testing strategies, quality checklists", "Standardized agent communication protocol based on agentic-interaction.json standard (message types: handoff, question, blocker, update, validation, completion)", "Context persistence in external database (separate from project files)", "Work order/task ID tracking system for completed work", "Agent summarization and logging capabilities", "Integration with archived projects", "Agent memory/context persistence across sessions", "Ability to define custom agent personas via config files", "Agent collaboration logs (track which agent did what)", "Must integrate seamlessly with existing docs-mcp server", "Must use MCP protocol (no custom protocols)", "Must work with Claude/Anthropic models primarily", "Must maintain backward compatibility with current docs-mcp tools" ], "out_of_scope": [ "Training custom AI models (use existing LLMs with specialized prompts) - FUTURE", "Real-time collaboration between multiple human users - FUTURE", "Integration with external project management tools (Jira, Asana, etc.) - FUTURE", "Agents that execute code autonomously without approval - FUTURE", "Multi-project support (focus on single project at a time) - FUTURE", "Web UI/dashboard (CLI/MCP interface only) - FUTURE" ], "constraints": [ "Implementation is flexible - no major technical, timeline, or budget constraints", "Should follow existing docs-mcp patterns and architecture", "Must adopt standardized agent communication protocol from agentic-interaction.json" ], "communication_protocol": { "source": "C:\\Users\\willh\\Desktop\\projects - current-location\\coderef-system\\coderef\\working\\coderef2-full-implementation\\agentic-interaction.json", "message_types": { "handoff": "Phase completion and transition to next agent", "question": "Request for clarification or guidance", "blocker": "Critical issue preventing progress", "update": "Progress update or status change", "validation": "Request for deliverable validation", "completion": "Phase completion notification" }, "message_structure": { "id": "Unique message ID (e.g., msg_001)", "timestamp": "ISO 8601 timestamp", "from": "Sending agent role/id", "to": "Receiving agent role/id", "phase": "Phase number or identifier", "type": "Message type (handoff, question, blocker, etc.)", "priority": "high | medium | low", "subject": "Brief message subject", "message": "Detailed message content", "status": "pending_acknowledgment | posted | acknowledged", "context_files": "Optional: Relevant context file paths", "action_required": "Optional: Specific actions needed" }, "priority_levels": { "high": "Important, should be addressed within same session", "medium": "Should be addressed before phase completion", "low": "Informational, no immediate action required" }, "response_expectations": { "handoff": "Acknowledge within same session, begin work immediately", "question": "Respond within same session if coordinator available", "blocker": "Respond immediately, resolve before continuing", "validation": "Validate within same session if possible" }, "handoff_protocol_steps": [ "Previous agent: Mark phase deliverables complete", "Previous agent: Create handoff message in tracking file", "Coordinator: Validate deliverables against success criteria", "Coordinator: Update phase_status and current_phase", "Coordinator: Post handoff message to next agent", "Next agent: Acknowledge message, update status to 'active'", "Next agent: Begin work on assigned phase" ], "coordination_rules": { "sequential_execution": "Only one phase active at a time (unless parallel work authorized)", "dependency_enforcement": "Phase N cannot start until Phase N-1 is complete and validated", "documentation_mandatory": "All decisions, learnings, and results must be documented", "commit_before_handoff": "All changes must be committed before phase handoff", "coordinator_validation": "Coordinator must validate deliverables before approving handoff" } }, "persistence_requirements": { "database_storage": "External database for agent context (separate from project files)", "work_order_tracking": "Task/work order IDs that persist outside project context", "context_summarization": "Agents can summarize and update context after task completion", "archived_project_integration": "Link agent work to archived projects", "session_persistence": "Agent memory persists across sessions", "collaboration_logs": "Track which agent performed which actions" }, "agent_roster": { "project_leader": { "type": "coordinator", "responsibilities": [ "Break down complex requests into subtasks", "Delegate tasks to appropriate specialist agents", "Monitor progress and coordinate between agents", "Make architectural decisions and approvals", "Validate deliverables against success criteria", "Ensure phase dependencies are met", "Coordinate handoffs between implementation agents", "Maintain documentation and changelog" ] }, "project_assistant": { "type": "assistant", "responsibilities": [ "Answer questions about the project (codebase, architecture, decisions)", "Track tasks and progress (TODO management, status updates)", "Run routine checks (tests, builds, linting)", "Gather information and do research", "Take notes and document decisions", "Work directly with user as interactive assistant" ] }, "specialist_agents": { "note": "Multiple specialist agents to be defined covering: UI/UX, Framework-specific, Backend, Testing, Documentation, Code Review, DevOps, Security, etc." } } }