apl-marketplace

APL - Autonomous Phased Looper

The ultimate autonomous coding plugin for Claude Code.

Set a goal. APL handles the rest - planning, coding, testing, reviewing, and learning for next time.

---

Quick Start

Install

# Add the marketplace
/plugin marketplace add https://github.com/twofoldtech-dakota/apl.git

# Install APL
/plugin install apl-autonomous-phased-looper@apl-marketplace

Use

/apl Build a REST API with JWT authentication

Monitor (Optional)

/apl gui

Opens a web dashboard at http://localhost:5173

---

What It Does

You: /apl Build a healthcare patient portal

APL: ┌──────────┐    ┌──────────┐    ┌─────────┐    ┌─────────┐
     │ DETECT   │ ─▶ │   PLAN   │ ─▶ │ EXECUTE │ ─▶ │ REVIEW  │
     └──────────┘    └──────────┘    └─────────┘    └─────────┘
          │               │               │              │
          ▼               ▼               ▼              ▼
     Auto-detect     Generate        Code each      Verify all
     complexity,     .apl/plan.md    story using    criteria,
     ask critical    with progress   ReAct loops    extract
     questions       checkboxes      + tests        learnings

Zero-config. Planning flows directly into execution. Progress tracked via checkboxes.

Then remembers what worked for next time.

---

Features

Feature	Description
Autonomous Execution	Plans, codes, tests, and reviews without hand-holding
26 Specialized Agents	Full digital team: coding, design, content, QA, analytics, research, docs, perf
Horizontal Quality Agents	Content strategy, brand voice, design, accessibility, copy/content
Enterprise Scale	Epic/Feature/Story hierarchies with autopilot mode
Self-Learning	Remembers successful patterns and avoids past failures
Web Dashboard	Visual monitoring and control via /apl gui
Error Recovery	Automatic retry with graduated fallback strategies
Checkpoints	State saves at each phase with rollback support

---

GUI Dashboard

APL includes a web-based control panel for visual monitoring.

/apl gui

URL	Service
http://localhost:5173	Dashboard
http://localhost:3001	API

Dashboard Features:

• Real-time workflow monitoring
• Phase and task progress visualization
• Agent activity tracking
• ReAct loop visualization
• Token usage monitoring
• Configuration editor
• Learnings browser
• Checkpoint management

---

Commands

Core APL Commands

Command	Description
/apl <goal>	Run APL with a coding goal
/apl gui	Launch web dashboard
/apl status	Show current state
/apl config	View configuration
/apl reset	Clear state, start fresh
/apl rollback <id>	Restore checkpoint
/apl forget <id>	Remove learned pattern

Enterprise Commands

Command	Description
/apl <goal>	Plan enterprise-scale project (auto-detects complexity)
/apl loop	Execute next Epic
/apl autopilot	Execute ALL remaining Epics continuously
/apl answer <id> <text>	Answer clarifying question

---

Examples

Single Session Tasks

# Build features
/apl Build a REST API with rate limiting and JWT auth

# Refactor
/apl Refactor the data layer to use repository pattern

# Debug
/apl Fix the race condition in order processing

# Test
/apl Add unit tests for all services with 80% coverage

# Migrate
/apl Migrate from Express to Fastify, keep API compatible

Enterprise Projects

# Plan a large project (APL auto-detects complexity)
/apl Build a healthcare patient portal with appointments and medical records

# Run through all Epics automatically
/apl autopilot

# Or step through manually
/apl loop          # Epic 1
/apl loop          # Epic 2
/apl loop          # Epic 3

---

How It Works

Phase 1: Plan

The Planner agent breaks down your goal:

• Creates atomic, verifiable tasks
• Defines success criteria
• Identifies dependencies and parallel opportunities
• Checks learned patterns from past sessions

Phase 2: Execute

The Coder agent implements each task using ReAct:

1. Reason - Analyze what's needed
2. Act - Write/modify code
3. Observe - Check results, run tests
4. Verify - Confirm success criteria met

On failure: retry → adjust approach → backtrack → escalate to user

Phase 3: Review

The Reviewer agent validates everything:

• Cross-task analysis
• Final success criteria check
• Learning extraction for future sessions

---

Self-Learning

APL improves with each session. All state is auto-managed in .apl/:

.apl/
├── plan.md           # Human-readable: progress with checkboxes
├── state.json        # Internal: machine state
├── learnings.json    # Internal: accumulated patterns
└── checkpoints/      # Internal: recovery points