# Ralph Orchestrator

A hat-based orchestration framework that keeps AI coding agents looping through planning and implementation until a task is done, with a web dashboard and per-repository MCP server.

Ralph Orchestrator is a hat-based orchestration framework that keeps AI agents in a loop until a task is complete, built as an improved implementation of the Ralph Wiggum technique for autonomous agent orchestration. It initializes with a chosen backend such as Claude, plans a feature through an interactive session that produces requirements, design and implementation-plan documents, and then implements the plan, iterating until it reports completion or hits an iteration limit; simpler tasks can skip planning and run directly.

An alpha web dashboard, started with its own command, monitors and manages orchestration loops through a Rust RPC API and a frontend that opens automatically in a browser, with configurable ports for both. Its MCP server is scoped to a single workspace root per instance, so multiple repositories each run their own server instance, with the workspace root resolved from a command-line flag, an environment variable, or the current directory in that order.

It is written in Rust, installs via npm, Cargo or a release installer script, and is released under the MIT license.

Website: https://github.com/mikeyobrien/ralph-orchestrator