Skip to content

Quick Start

Get up and running with the AI Context System in 10 minutes.

Installation (2 minutes)

One-Command Install

bash
curl -sL https://raw.githubusercontent.com/rexkirshner/ai-context-system/main/install.sh | bash

What this does:

  • Downloads system files to current directory
  • Creates .claude/, scripts/, templates/ folders
  • Installs 14 slash commands
  • Sets up configuration

Verify Installation

bash
# Check if .claude directory exists
ls -la .claude/commands/

# Should see:
# init-context.md
# save.md
# save-full.md
# review-context.md
# ... and 10 more commands

Initialize Context (3 minutes)

Run /init-context

bash
/init-context

What happens:

  1. Creates context/ directory
  2. Generates 5 core files:
    • claude.md - Entry point for AI
    • CONTEXT.md - Project orientation
    • STATUS.md - Current state
    • DECISIONS.md - Decision log
    • SESSIONS.md - Session history
  3. Creates .context-config.json - Project configuration

Fill Out CONTEXT.md

Edit context/CONTEXT.md with your project details:

markdown
## What & Why

**Project:** [Your project name]
**Business Goal:** [What problem does it solve?]
**Target Users:** [Who uses it?]

## Tech Stack

**Frontend:** [e.g., Next.js 14, React, TypeScript]
**Backend:** [e.g., PostgreSQL, Express, Prisma]
**Infrastructure:** [e.g., Vercel, AWS, Docker]

Time: 3-5 minutes to fill in the basics

Your First Session (5 minutes)

1. Check Current Status

bash
/review-context

What you see:

  • Quick Reference (project overview)
  • Current work in progress
  • Last session summary
  • System version check

2. Start Working

Do some actual work on your project:

  • Write code
  • Fix bugs
  • Add features
  • Refactor
  • Whatever you're working on

3. Save Your Progress

Every 30-60 minutes:

bash
/save

What this does:

  • Updates STATUS.md with current work
  • Regenerates Quick Reference
  • Captures work in progress
  • Takes 2-3 minutes

4. Before Break - Comprehensive Save

At end of session (lunch, end of day):

bash
/save-full

What this does:

  • Everything /save does PLUS:
  • Creates new SESSIONS.md entry
  • Documents mental models
  • Captures decision rationale
  • Auto-logs git operations
  • Takes 10-15 minutes

Daily Workflow

bash
# 1. Start of day
/review-context

# 2. Work on project
#    (code, debug, refactor, etc.)

# 3. Save frequently (every 30-60 min)
/save

# 4. Before lunch break
/save-full

# 5. More work after lunch
#    (more coding)

# 6. Save frequently
/save

# 7. End of day
/save-full

Example First Session

Here's what a complete first session looks like:

1. Install (9:00 AM)

bash
cd ~/projects/my-app
curl -sL https://raw.githubusercontent.com/rexkirshner/ai-context-system/main/install.sh | bash

2. Initialize (9:02 AM)

bash
/init-context

Edit context/CONTEXT.md:

markdown
## What & Why

**Project:** Task Management App
**Business Goal:** Help teams organize work asynchronously
**Target Users:** Remote engineering teams (10-50 people)

## Tech Stack

**Frontend:** Next.js 14, TypeScript, Tailwind
**Backend:** PostgreSQL, Prisma
**Infrastructure:** Vercel

3. Start Session (9:10 AM)

bash
/review-context

# Shows:
# - Project: Task Management App
# - Phase: Initial Setup
# - No work in progress yet

4. Work - Build Authentication (9:10 AM - 11:00 AM)

  • Set up Supabase
  • Create auth pages (login, signup)
  • Implement JWT session handling
  • Add protected routes

5. First Save (11:00 AM)

bash
/save

AI updates STATUS.md:

markdown
## Work In Progress

**Current Focus:** Authentication setup

- ✅ Supabase project created
- ✅ Auth pages (login/signup) built
- 🔄 JWT session handling (in progress)
- ⏳ Protected routes (next)

6. Keep Working (11:00 AM - 12:30 PM)

  • Finish JWT handling
  • Add protected routes
  • Test auth flow

7. Lunch Break - Full Save (12:30 PM)

bash
/save-full

AI creates SESSIONS.md entry:

markdown
## Session 1 | 2025-10-24 | Initial Authentication Setup

**TL;DR:** Set up Supabase auth, built login/signup pages,
implemented JWT session handling, added protected routes.
Auth flow working end-to-end.

### Mental Models

**Current understanding:**
- Using Supabase for authentication (email + OAuth)
- JWT tokens in HTTP-only cookies
- Protected routes check session validity
- Next.js middleware handles route protection

**What's working:**
- ✅ User can sign up with email
- ✅ User can log in
- ✅ Session persists across page reloads
- ✅ Protected routes redirect to login

### Next Session

Start here:
1. Add password reset flow
2. Implement OAuth providers (Google, GitHub)
3. Test error cases (invalid email, network errors)

8. Resume After Lunch (1:30 PM)

bash
/review-context

# Sees:
# - Last session: Authentication setup
# - Next steps: Password reset, OAuth
# - Current work: Auth complete, ready for next feature

Continue with password reset flow...

9. End of Day (5:00 PM)

bash
/save-full

# Creates Session 2 entry
# Documents password reset implementation
# Captures decision about email service
# Auto-logs git commits

Next Day

bash
# 9:00 AM next day
/review-context

# Immediately see:
# - Project state
# - What was done yesterday
# - Where to continue
# - Any blockers

# Resume work with full context in 2 minutes
# No re-explaining needed
# Zero context loss

Key Commands Reference

CommandWhenTimeWhat It Does
/init-contextOnce (setup)3 minCreates context structure
/review-contextStart of session2 minShows current state + Quick Reference
/saveEvery 30-60 min2-3 minQuick STATUS.md update
/save-fullBefore breaks10-15 minComprehensive save + SESSIONS.md
/validate-contextBefore handoffs2 minCheck documentation health
/export-contextFor handoffs1 minPackage everything for sharing

Tips for Success

1. Save Often

bash
# Don't wait until end of day
# Save every 30-60 minutes
/save

# Safety net for unexpected interruptions
# Captures incremental progress

2. Full Saves at Boundaries

bash
# Before any break:
/save-full

# Lunch, meetings, end of day, context switches
# Comprehensive state capture

3. Review at Start

bash
# First thing every session:
/review-context

# Orientation in 2 minutes
# Know where to start
# No context loss

4. Trust the System

markdown
# Document even "obvious" decisions
# Capture mental models honestly
# Future you will thank present you

# Example:
## DEC-001: Use PostgreSQL

**Decision:** PostgreSQL for database

**Why:** Need ACID transactions, complex relationships

# Even if "obvious", document it
# New AI agent needs to know

5. Read the Documentation

After first session, read:

Troubleshooting

Commands Not Working?

Problem: Slash commands don't work

Solution:

bash
# Check for multiple .claude directories
find ~ -name ".claude" -type d

# Should only have one in your project root
# If multiple found, remove extras

# Keep only:
/path/to/your-project/.claude

Can't Find Context Files?

Problem: Commands can't find context/ directory

Solution:

bash
# Make sure you're in project root
pwd

# Should see:
/path/to/your-project

# Not a subdirectory like:
/path/to/your-project/src  # Wrong!

The system auto-detects context up to 3 parent directories, but works best from project root.

Quick Reference Not Generated?

Problem: STATUS.md Quick Reference section is empty

Solution:

bash
# Check .context-config.json exists
ls context/.context-config.json

# If missing, run:
/init-context

# Then run:
/save

Success Checklist

After first session, verify:

  • [ ] context/ directory exists
  • [ ] context/CONTEXT.md filled out with project details
  • [ ] context/STATUS.md shows current work
  • [ ] context/SESSIONS.md has Session 1 entry
  • [ ] Quick Reference auto-generated in STATUS.md
  • [ ] /review-context shows project state
  • [ ] /save updates STATUS.md successfully
  • [ ] /save-full creates SESSIONS.md entry

If all checked: You're ready to use the system! 🎉

Next Steps

Learn the concepts:

Explore commands:

Learn workflows:

Remember

"When in doubt, /save!"

The system is your safety net. Save frequently, trust the process, and enjoy perfect session continuity.

Built for all AI coding assistants. Optimized for Claude Code.