
Ralph Wiggum Loop
io.github.sheldonrobinsonPublished by GitHub account sheldonrobinsonv1.1.0Updated Oct 27, 2026
Autonomous iterative AI development MCP server with spec-driven worker/reviewer SHIP/REVISE loops.
Overview
Local MCP server that runs autonomous iterative development loops, with a worker model implementing a spec and a reviewer model deciding SHIP or REVISE.
- What you can do
- Run a two-phase iterative development loop where a worker model does the work and a reviewer model reviews it until it says SHIP.
- Drive the loop from the command line with a task prompt and a profile, or expose it to an assistant over stdio or streamable HTTP.
- Load workflow recipes from a configured recipe directory and keep loop state and iteration history in a configured state directory.
- Install as an agent plugin with bundled skills for spec-driven development and stop hooks.
- When to use it
- Useful when you want an assistant to carry out a multi-step coding task autonomously and have a second model review each iteration before the work is accepted. Suited to spec-driven development work on a local machine rather than one-off questions.
- Requirements
- Runs as a local process on the user's machine (desktop only). Needs a POSIX shell on Linux/macOS or PowerShell on Windows; the HTTP mode listens on a local port such as 127.0.0.1:8080. Two directories must be configured through the RALPH_RECIPE_DIR and RALPH_STATE_BASE environment variables. No authentication is declared.
Installation
In SourceWeft
- Open Ralph Wiggum Loop in the dashboard and add it to a workspace.
- Enable the server for the chats that should use its tools.
Desktop only via STDIO. STDIO servers start a local process, so they need the SourceWeft desktop host.
Other MCP clients
Follow the launch instructions in the repository.
README
ralph-wiggum-loop
Multi-model iterative development agent plugin — Ralph Loop with cross-model worker/reviewer architecture, spec-driven development skills, Stop hooks, and SHIP/REVISE decision engine.
[License: MIT] [Agent Plugin] [GitHub Pages]
Overview
The Ralph Loop implements a two-phase iterative workflow where a worker model does the work and a reviewer model provides cross-model review, iterating until the reviewer says "SHIP".
This project is a cross-platform Agent Plugin (plugin.json, mcp.json, .agents/skills/ralph-wiggum-loop/ & skills/ralph-wiggum-loop/) with native implementations for:
- Linux / macOS (Bash):
bin/ralph-wiggum-loop.sh(usingbin/mcp-sdk.sh) - Windows (PowerShell):
bin/ralph-wiggum-loop.ps1(usingbin/mcp-sdk.psm1) - Polyglot Runner:
bin/script-runner.sh/bin/script-runner.ps1(runs in both POSIXsh/bashand PowerShell)
Install Skill (.agents/skills/)
Per the Agent Skills Specification, install the skill to .agents/skills/:
[Ralph Wiggum Loop MCP server]
OS-Specific MCP Servers (mcp.json)
Quick Start
1. CLI Mode
- Linux / macOS (Bash):
- Windows (PowerShell):
2. MCP Server (stdio Mode)
- Linux / macOS (Bash):
- Windows (PowerShell):
3. MCP Server (streamable-http Service Mode)
- Linux / macOS (Bash):
- Windows (PowerShell):
Documentation
- Skill & Agent Instructions — OS selection rules, loop execution modes, and Streamable HTTP setup
- Full Documentation — API reference, tools, state management
- Deployment Guide — Install and configure
Repository
- Source: https://github.com/sheldonrobinson/ralph-wiggum-loop
- Agent Plugin Spec: https://agent-plugins.org
- MCP Spec: https://modelcontextprotocol.io
License
MIT
Source: README.md at commit f094b28
Tools
0Version history
1- v1.1.0LatestOct 23, 2026
Score
50% of the applicable checks met
How complete this listing is, computed from the registry entry, the endpoint and public sources. Not a rating of the server and not a statement that it is safe or working today: reachability is a dated check, and no installs or reviews go into the score.
- Reached in the last 30 days25 pts
Keep the endpoint answering the MCP handshake; SourceWeft probes it every 30 days.
- Auth declared10 pts
Declare the credential headers in the registry entry, or keep the endpoint answering the probe so its auth can be classified.
- Tools listed15 pts
Answer tools/list without a credential, or ship a package that starts in a sandbox without one.
- DescriptionRequired15 pts
- Logo10 pts
- Documentation15 pts
- License10 pts