alfredolopez80/multi-agent-ralph-loop

Multi-Agent Ralph Loop

Autonomous orchestration framework for Claude Code with MemPalace-inspired memory, parallel-first Agent Teams, Aristotle First Principles methodology, and quality gates.

What It Does

Ralph extends Claude Code into a multi-agent development framework with a structured memory system inspired by the Memory Palace technique. Every task is analyzed from first principles, decomposed into parallel subtasks, assigned to specialized teammates, and validated through quality gates before completion.

| Capability | Description | |---|---| | MemPalace Memory | 4-layer memory stack (L0-L3) with Obsidian vault knowledge graph and learned rules taxonomy | | Parallel-First | All independent tasks execute in parallel via Agent Teams (mandatory for complexity >= 3) | | 6 Teammates | ralph-coder, ralph-reviewer, ralph-tester, ralph-researcher, ralph-frontend, ralph-security | | Hook System | Lifecycle hooks for validation, quality gates, security guards, and automatic learning | | Aristotle Analysis | 5-phase first principles deconstruction before every non-trivial task | | Quality Gates | 4-stage blocking validation: correctness, quality, security, consistency | | Comprehensive Tests | Full test suite covering layers, hooks, security, skills, and pipeline |

MemPalace Memory System

Inspired by the MemPalace repository (Memory Palace technique for LLM agents), Ralph implements a layered memory architecture with key differences based on our implementation findings.

Layer Stack (Session Wake-up)

| Layer | File | Purpose | |-------|------|---------| | L0 | ~/.ralph/layers/L0_identity.md | Agent identity + principles | | L1 | ~/.ralph/layers/L1_essential.md | Actionable rules (filtered from corpus) | | L2 | .claude/rules/learned/{halls,rooms,wings}/ | Project-specific taxonomy (on-demand) | | L3 | Obsidian vault grep | Full kno