Quick Start Guide (v2.0)

Get up and running with Mini-CoderBrain v2.0 in 2 minutes with intelligent setup!

🚀 Step 1: Install (30 seconds)

git clone https://github.com/kunwar-shah/mini-coder-brain.git
cd mini-coder-brain
chmod +x install.sh
./install.sh /path/to/your/project

Installation complete! But there’s one more critical step…

🔴 Step 2: Initialize Context (MANDATORY)

⚠️ CRITICAL: This step is REQUIRED for mini-coder-brain to work!

Open your project in Claude Code and run ONE of these:

Option A: With Documentation (Best!)

/init-memory-bank --docs-path ./docs

Use this if you have:

  • SRS.md, ARCHITECTURE.md, API.md
  • Any technical documentation
  • Requirements documents

Result: 90-95% context quality ⭐

Option B: Auto-Detection

/init-memory-bank

Use this if you have:

  • Existing code (package.json, README.md)
  • Git history
  • No formal documentation

Result: 75-85% context quality ✅

Option C: Interactive Wizard

/init-memory-bank

Use this if you have:

  • Brand new empty project
  • Just starting

Result: 60-70% context quality (minimum) ⚠️

✅ Step 3: Verify Setup

Check your context quality:

/validate-context

Expected output:

📊 Context Quality: 85% (Recommended) ✅

✅ Project name: your-project
✅ Tech stack: 5 technologies
✅ Core features: 4 features
✅ Architecture: Defined

✅ Ready for development!

If score < 60%: Run /validate-context --fix for help.

🎯 Step 4: Start Developing!

Now you can just talk to Claude naturally:

"Add JWT authentication to the user login endpoint"

Claude will:

  • ✅ Know your project structure automatically
  • ✅ Follow your coding patterns
  • ✅ Use your exact tech stack
  • ✅ Remember all decisions
  • ✅ Track progress across sessions
  • ✅ Never ask “what framework are you using?” again!

🧠 Understanding the Memory Bank

Mini-CoderBrain maintains 5 key memory files:

1. productContext.md - Project Overview

What: Project name, tech stack, architecture, features When to edit: Once during setup, update when scope changes

2. activeContext.md - Current Focus

What: What you’re working on RIGHT NOW When to edit: Updated automatically by hooks, or use /context-update

3. progress.md - Sprint Tracking

What: Completed/in-progress/upcoming tasks When to edit: Updated automatically, or manually for milestones

4. decisionLog.md - Technical Decisions

What: ADRs (Architecture Decision Records) with timestamps When to edit: Automatically appended when making technical decisions

5. systemPatterns.md - Coding Standards

What: Your coding patterns, conventions, testing approach When to edit: Once during setup, update when standards change

🛠️ Essential Commands

Memory Management

/memory-sync              # Full memory synchronization
/context-update          # Quick real-time context update
/umb "note"             # Fast manual sync with note
/memory-cleanup         # Archive old data, prevent bloat

Codebase Mapping

/map-codebase           # Load existing map (instant)
/map-codebase --rebuild # Rebuild from scratch (~30s)
/map-codebase --recent  # Focus on recent changes only

💡 Common Workflows

Starting a New Feature

  1. Set your focus: 
    /context-update focus "Building user profile page"
  2. Start coding: 
    "Create a user profile component with avatar, bio, and social links"
  3. Claude automatically:
    • Uses your project structure
    • Follows your coding patterns
    • Tracks this work in activeContext
    • Updates progress

Making Technical Decisions

Just discuss with Claude naturally:

"Should we use Redux or Context API for state management?"

Claude will:

  • Consider your tech stack
  • Reference past decisions
  • Provide recommendations
  • Automatically log the decision to decisionLog.md

Ending a Session

Just close Claude Code. The stop hook automatically:

  • ✅ Saves session summary to activeContext.md
  • ✅ Tracks development operations
  • ✅ Suggests memory cleanup if needed
  • ✅ Preserves all context for next session

🎯 Best Practices

DO ✅

  • Use /context-update to keep Claude informed of your focus
  • Run /map-codebase once for instant file access
  • Let hooks work automatically - they track everything
  • Trust the memory bank - Claude reads it at session start
  • Run /memory-cleanup when notified (prevents bloat)

DON’T ❌

  • Don’t manually edit memory files during development (hooks do it)
  • Don’t skip /memory-cleanup notifications (causes “Prompt too long”)
  • Don’t delete .claude/tmp/ while Claude is running
  • Don’t commit .claude/memory/*.md files (they’re gitignored)

📊 Status Footer

Every Claude response ends with:

🧠 MINI-CODER-BRAIN STATUS
📊 Activity: X ops | 🗺️ Map: Status | ⚡ Context: Active

This shows:

  • Activity: Operations performed this session
  • Map: Codebase map status (None/Loaded/Available)
  • Context: Context awareness status

🔄 Session Workflow

Session Start

  1. Hooks load context from .claude/memory/
  2. Project structure detected automatically
  3. Boot status displays current focus
  4. You’re ready to code with full context

During Development

  1. Work naturally with Claude
  2. Context tracks automatically
  3. Progress updated in real-time
  4. Decisions logged automatically

Session End

  1. Stop hook saves session summary
  2. Context preserved for next session
  3. Smart suggestions if cleanup needed
  4. Perfect continuity next time

🆘 Quick Troubleshooting

“Claude doesn’t remember my project”

Solution: Check that CLAUDE.md and .claude/memory/ files exist

“Prompt is too long” error

Solution: Run /memory-cleanup to archive old data

“Commands don’t work”

Solution: Verify hooks are executable: chmod +x .claude/hooks/*.sh

🎓 Advanced Features

Codebase Mapping

Revolutionary instant file access:

  • One-time rebuild: 30 seconds
  • Future loads: instant
  • Surgical precision: Claude knows exact file locations

Real-Time Context Updates

Track focus/blockers/achievements:

/context-update achievement "Completed authentication system"
/context-update blocker "API rate limiting issue"

Memory Cleanup System

Prevents “Prompt too long” errors:

  • Automatic suggestions when needed
  • Archives old data (keeps last 5 session updates)
  • No data loss - everything preserved in .claude/archive/

📚 Next Steps

Now that you’re set up:

  1. Customize your memory bank - Edit productContext.md and systemPatterns.md
  2. Map your codebase - Run /map-codebase --rebuild once
  3. Start developing - Claude has perfect context awareness!
  4. Explore features - Check out the Features Guide
  5. Learn commands - See Commands Reference

💬 Example Conversation

You: “Add user authentication with JWT tokens”

Claude (with Mini-CoderBrain):

  • ✅ Checks productContext.md → sees you’re using Express + MongoDB
  • ✅ Checks systemPatterns.md → sees you use bcrypt for passwords
  • ✅ Checks project-structure.json → finds src/routes/auth.js
  • ✅ Creates authentication following YOUR patterns
  • ✅ Updates activeContext with current work
  • ✅ No questions needed - just starts coding!

Without Mini-CoderBrain:

  • ❌ “What framework are you using?”
  • ❌ “Where should I create the auth routes?”
  • ❌ “What’s your preferred password hashing library?”
  • ❌ Loses context next session
  • ❌ Repeats questions every time

Ready to explore all features?

View All Features →