What This Stack Does
This stack transforms AI-assisted development from single-agent conversations into a coordinated multi-agent workflow with persistent memory and intelligent code retrieval. Oh My ClaudeCode adds 32 specialized sub-agents inside each Claude Code session — an architect, executor, QA tester, security sentinel, and more — with smart model routing that saves 30-50% on token costs. Baton manages multiple agent sessions in parallel across isolated git worktrees, with a visual diff viewer and automatic PR creation. Supermemory ensures every agent session has access to your accumulated project context. LightRAG builds a knowledge graph from your documentation that agents can query for relationship-aware retrieval.
The Bottom Line
The workflow leverages each tool's strengths in sequence. Start by indexing your project documentation and codebase into LightRAG to build a searchable knowledge graph. Configure Supermemory's MCP server so all your AI tools share persistent memory about your project architecture, coding conventions, and past decisions. Launch Baton and create multiple workspaces — each running Claude Code with Oh My ClaudeCode installed. Assign different features or refactoring tasks to each workspace. OMC's specialized agents handle architecture design, implementation, testing, and documentation within each session, while Baton keeps the work isolated across git worktrees. When agents need context about how your codebase is structured, LightRAG provides relationship-aware retrieval that goes beyond simple text search. The result is a development workflow where multiple AI teams work in parallel, each with full project context and specialized expertise.