PAUL Framework

1.2.0 · active · verified Sun Apr 19

PAUL (Plan-Apply-Unify Loop) is an AI-assisted development framework designed to bring structure and integrity to projects built using Claude Code. Currently at version 1.2.0, the framework emphasizes disciplined execution through a cyclical process: Plan, Apply, and Unify. It aims to combat "context rot" and improve the reliability of AI-generated output by enforcing loop integrity, prioritizing in-session context, and driving development with acceptance criteria. PAUL ensures that plans are closed, state is consistently managed, and decisions are logged, significantly reducing the debugging of AI output. It is not limited to software development but can manage diverse projects like marketing campaigns and workflows. Releases are driven by user feedback, focusing on enhancing workflow quality and depth, with the latest v1.2.0 introducing significant improvements to the initialization process.

Common errors

Warnings

Install

Imports

Quickstart

Demonstrates the initial setup of PAUL via `npx` and the core Plan-Apply-Unify loop commands used within Claude Code to manage an AI-assisted project.

npx paul-framework

# Choose location (Global or local)
# Verify installation inside Claude Code:
/paul:help

# Quick Workflow inside Claude Code:
# 1. Initialize PAUL in your project with a type-adapted requirements conversation
/paul:init

# 2. Create a plan for your work, auto-detecting scope and validating coherence
/paul:plan

# 3. Execute the approved plan, with each task going through Execute/Qualify loop
/paul:apply

# 4. Close the loop, reconciling planned vs. actual outcomes and updating state
/paul:unify

# 5. Check progress anytime
/paul:progress

view raw JSON →