Ralph Orchestrator

Active
GitHub Rust MIT

Description

An improved implementation of the Ralph Wiggum technique for autonomous AI agent orchestration, built in Rust for reliable multi-agent task coordination and scheduling.

Key Features

  • Ralph Wiggum autonomous iteration technique - Keeps AI agents in a continuous loop until task completion or iteration limit
  • Hat system - Specialized personas coordinating through events with built-in patterns (code-assist, debug, research, review, pdd-to-code-assist)
  • Multi-backend support - Compatible with Claude Code, Kiro, Gemini CLI, Codex, Amp, Copilot CLI, OpenCode
  • Backpressure gates - Rejects incomplete work via test, lint, and typecheck gates to ensure code quality
  • Persistent memories and tasks - Cross-session continuous learning and runtime task tracking
  • Telegram human-in-the-loop (RObot) - Bidirectional interaction via Telegram with agent question blocking and proactive guidance

Use Cases

💡 Let AI agents autonomously iterate on features until all quality gates pass (tests, lint, typecheck)
💡 Coordinate multiple specialized roles in a single workflow using the Hat system from planning to implementation
💡 Use Ralph Plan for interactive requirement planning, generating requirements, design, and implementation-plan docs
💡 Enable remote human-in-the-loop collaboration via Telegram with agent-initiated guidance requests

Tags

multi-agent rust agent automation

Categories

🧠 Multi-Agent 🔄 Workflow

Quick Start

Install: npm install -g @ralph-orchestrator/ralph-cli. Initialize: ralph init --backend claude. Plan: ralph plan 'Add user authentication with JWT' (interactive PDD session, generates specs/). Implement: ralph run -p 'Implement the feature in specs/user-authentication/'. Simple tasks can skip planning with ralph run -p '...'.
Visit GitHub

Related Projects