Home

Emergent Learning Framework Wiki

Welcome to the ELF documentation. This wiki contains detailed guides for all features.

Quick Links

Page	Description
Installation	Prerequisites, install options, troubleshooting
Configuration	CLAUDE.md, settings.json, hooks setup
Dashboard	Visual monitoring, tabs, themes
Checkin	/checkin command - load context at session start
Checkout	/checkout command - record learnings before ending session
Swarm	Multi-agent coordination, personas
CLI Reference	All query commands and slash commands
Golden Rules	Customizing constitutional principles
Migration	Upgrading, team setup, rollback
Architecture	Database schema, hooks system
Token Costs	Usage breakdown, optimization tips

The Core Concept

Claude Code is stateless - every session starts from zero. This framework bridges that gap by giving Claude access to your project's history:

• Before each task: Use /checkin to query the building for relevant knowledge
• After each task: Use /checkout to record outcomes automatically
• Over time: Patterns strengthen, mistakes don't repeat

The Learning Loop

QUERY   ->  /checkin loads knowledge from building
APPLY   ->  Use heuristics during task
RECORD  ->  /checkout captures learnings
PERSIST ->  Update confidence scores
            |
         (cycle repeats)

What It's NOT

• Not AI training - Claude's model doesn't change. This is structured context injection.
• Not magic - It's systematic note-taking with automatic retrieval.
• Not instant value - It compounds over time. Day one won't feel different.

Getting Started

1. Install: Follow Installation
2. Configure: Set up Configuration
3. First session: Run /checkin to load context
4. During work: Use golden rules and heuristics
5. End of session: Run /checkout to record learnings
6. Next session: See what you learned in /checkin

The learning cycle (/checkin → work → /checkout) is the heart of ELF.