multi-agent-ralph-loop and smart-ralph
About multi-agent-ralph-loop
alfredolopez80/multi-agent-ralph-loop
Orchestration system for Claude Code with memory-driven planning, multi-agent coordination, Agent Teams integration, automatic learning, and comprehensive security validation (Grade A-). v2.94.0
Implements multi-agent orchestration through 13 MCP servers (ast-grep for structural code search, Zai vision tools, web search) with parallel memory backends, episodic TTL-based cleanup, and task routing to Claude Code instances via swarm mode. Integrates GitHub repository learning, LSP-based code navigation (TypeScript, Python, C/C++), and adversarial security validation across 950+ BATS tests achieving A- grade compliance.
About smart-ralph
tzachbon/smart-ralph
Spec-driven development with smart compaction. Claude Code plugin combining Ralph Wiggum loop with structured specification workflow.
Executes development workflows through five sequential specification phases—research, requirements, design, tasks, and implementation—with context isolation per phase to prevent hallucination drift. Supports both Claude Code and Codex platforms via plugin/skill architecture with approval gates and quick-mode execution options. Maintains zero external dependencies while generating structured markdown artifacts (research.md, requirements.md, design.md, tasks.md) that persist across sessions.
Related comparisons
Scores updated daily from GitHub, PyPI, and npm data. How scores work