# ChristopherKahler/paul

Plan-Apply-Unify Loop — Structured AI-assisted development for Claude Code. Quality over speed-for-speed's-sake.

Repository: https://github.com/ChristopherKahler/paul
Canonical: https://ross.abutalabs.com/products/paul
Homepage: https://github.com/ChristopherKahler/paul/blob/main/PAUL-VS-GSD.md
Language: JavaScript
License: MIT
License Family: permissive
Topics: ai-development, ai-tools, claude, claude-code, development-framework, project-orchestration, spec-driven, vibe-coding
Last push: 2026-08-21T22:08:43+00:00

## Health v2 (maintenance only)
Score: 62/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 98, release rhythm 44, longevity 15
- inputs: {"age_days": 217, "days_push": 12, "days_rel": 162, "gap_med": null, "n_releases_24m": 1}
- flags: none
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1205, forks 128 (observed 2026-08-28T04:03:59.135238+00:00)

## What it is
PAUL (Plan-Apply-Unify Loop) is a structured development framework for Claude Code that enforces a PLAN → APPLY → UNIFY cycle with acceptance criteria, state persistence, and decision logging. It combats context rot by keeping implementation in-session and reserving subagents for research, distributed as an npm package runnable via npx.

## Use cases
- structure ai-assisted development with claude code
- prevent context rot in long ai coding sessions
- enforce acceptance criteria in ai-generated code
- manage spec-driven development workflows
- track planned vs completed work across ai sessions
- run structured non-code projects like marketing campaigns with ai

## When to choose
- you use Claude Code daily and suffer from drifting plans and degraded session quality
- you want enforced loop discipline with reconciliation (UNIFY) and audit trails
- you prefer in-session execution over subagent sprawl
- you want acceptance-criteria-first, BDD-style task qualification

## When to avoid
- you don't use Claude Code or another compatible AI coding agent
- you want lightweight ad-hoc prompting without structured state files
- your workflow depends on heavy subagent parallelization for implementation
- you need a battle-tested enterprise framework rather than a young community tool

## Facets
- artifact type: framework
- maturity: active
- function: agent-framework, prompt-engineering, workflow-automation, developer-tools
- domain: developer-tools, large-language-models
- platform: windows, cli
- tags: claude-code, spec-driven-development, vibe-coding, context-management, acceptance-criteria, plan-execute-loop, ai-assisted-development, automation, ai-agents, nodejs, macos, linux

## Member repositories
- ChristopherKahler/paul (main) score 62

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:03:59.135238+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T06:19:19.972511+00:00, confidence not recorded.
  - readme: https://github.com/ChristopherKahler/paul (fetched 2026-08-28T04:03:59.135238+00:00, sha 4bfd8a48d308)
  - homepage: https://github.com/ChristopherKahler/paul/blob/main/PAUL-VS-GSD.md (fetched 2026-08-29T12:27:15.876959+00:00, sha 426122f1d5aa)
- Data as of 2026-08-30T08:39:29.467469+00:00.