Relentless
Relentless
Universal AI Agent Orchestrator That Loops Until Done
Run any AI coding agent (Claude Code, Amp, OpenCode, Codex, Droid, Gemini) repeatedly in an autonomous loop until all tasks are complete. Fresh context each iteration. Memory via Git. Pure magic.
Because your AI agent shouldn't quit after one try. Relentless keeps going until the job is done.
Fresh agent instance each iteration with clean context. Memory persists via Git history, progress.txt, and prd.json. No context window limits.
Works with Claude Code, Amp, OpenCode, Codex, Droid, and Gemini. Three tiers of support from full skills to CLI-only workflows.
Automatically switches agents when rate limits hit. Auto-recovery when limits reset. Never stop working.
Interactive specification via /relentless.specify, technical planning, hierarchical task breakdown across 4 phases.
Validates story dependencies, prevents circular references, executes tasks in correct order automatically.
Real-time terminal interface with progress bars, story grid, and live agent output. Watch the magic happen.
Inspired by Geoffrey Huntley's brilliant idea: keep spawning fresh AI agents until the work is done. We took it further with multi-agent support, smart routing, and cost optimization.
According to Matt Pocock, if you're trying to sneak Ralph into the enterprise, just pretend it's an acronym:
Relentless Agentic Looping Programming Helper
So naturally, RELENTLESS is the recursive acronym:
RALPH Endlessly Loops, Executing, Navigating Tasks, Learning, Evolving, Solving, Scaling
It's recursive. We loop. That's kind of our thing.
Use your favorite AI coding agent. Or let Relentless pick the best one for each task.
A simplified 5-step SpecKit workflow. Tasks auto-convert to prd.json with smart routing.
Describe your feature in natural language. Interactive clarification ensures nothing is missed.
Generate technical implementation plan with architecture decisions and patterns.
Create dependency-ordered user stories. Auto-converts to prd.json with smart routing.
Validate consistency across all artifacts. Catch issues before implementation.
AI agents build with TDD. Loop until done. Smart routing saves 50-75% on costs.
We're building the future of AI-assisted development. Here's what's coming.
Run any of 6 supported AI agents with automatic fallback when rate limits hit. Fresh context each iteration.
Interactive specification, technical planning, task breakdown, and quality checklists. Full workflow support.
Automatically use cheaper/faster models (Haiku, Gemini Flash, GPT-5.2 low reasoning-effort) for simple tasks. Save 50-75% without sacrificing quality.
Track costs per feature. Four cost modes (free, cheap, good, genius). Automatic escalation when needed.
Run multiple agents simultaneously on independent stories. Faster feature development.
Monitor all your Relentless runs from a beautiful web interface. Team collaboration features.
Install Relentless and start automating your AI coding workflow in minutes.
# Install globally with bun (recommended)
bun install -g @arvorco/relentless
# Initialize in your project
cd your-project
relentless init
# Create spec and run (Claude Code / Amp)
/relentless.constitution # One-time setup
/relentless.specify Add user authentication with OAuth2
/relentless.plan
/relentless.tasks # Auto-converts to prd.json!
/relentless.analyze # Validate consistency
# Run with beautiful TUI
relentless run --feature 001-auth --tui
We're building the future of AI-assisted productivity. Relentless is open source, but it's just the beginning.