A Ralph Loop that installs Ralph Loops.
Fleet orchestrator for managing Ralph infrastructure across all your GitHub repos.
┌─────────────────────────────────────────────────────────────┐
│ ralph-fleet (this repo) │
│ │
│ AGENTS.md + prd.json + GitHub API access │
│ │
│ Commands: │
│ - scan → list repos, detect Ralph status │
│ - status → fleet overview dashboard │
│ - install → add Ralph infrastructure to repo │
│ - sync → push config updates across fleet │
└─────────────────────────────────────────────────────────────┘
│
▼
┌──────────┬──────────┬──────────┬──────────┐
│ repo-a │ repo-b │ repo-c │ repo-d │
│ │ │ │ │
│ AGENTS.md│ AGENTS.md│ (none) │ AGENTS.md│
│ prd.json │ prd.json │ │ prd.json │
└──────────┴──────────┴──────────┴──────────┘
│
▼
`install repo-c`
│
▼
repo-c now has
Ralph installed
- Standardize - Same Ralph infrastructure across all repos
- Orchestrate - One place to manage fleet configuration
- Scale - Install Ralph on 50 repos in one command
- Dogfood - This repo itself runs as a Ralph Loop
# Clone
git clone https://github.com/YOUR_USERNAME/ralph-fleet
cd ralph-fleet
# Setup
bun install
cp .env.example .env
# Add your GITHUB_TOKEN
# Scan your repos
bun run scan --days=30
# Install Ralph on a repo
bun run install owner/repo-name
# Install on all repos updated in last 30 days
bun run install --all --days=30When you run install, target repos receive:
repo/
├── AGENTS.md # Ralph instructions
└── scripts/
└── ralph/
├── prd.json # Task tracking
└── progress.txt # Learnings log
ralph-fleet itself uses the Ralph methodology:
AGENTS.md- Instructions for working on this reposcripts/ralph/prd.json- Stories to build the orchestratorscripts/ralph/progress.txt- What we've learned
To contribute or extend, just pick the next pending story from prd.json.
"As model quality, context, and speed race to commodity, the only moat left is the harness - your verification gates."
The loop compresses to nothing. The harness stays. Ralph Fleet helps you install that harness everywhere.
MIT