Claude Copilot - Quick Reference Card

Claude Copilot - Quick Reference Card

Installation Quick Commands

Machine Setup (Once Per Machine)

# Clone to ~/.claude/copilot
mkdir -p ~/.claude && cd ~/.claude
git clone https://github.com/Everyone-Needs-A-Copilot/claude-copilot.git copilot

# Open and run setup
cd ~/.claude/copilot && claude
# Then say: Read @~/.claude/copilot/SETUP.md and set up Claude Copilot on this machine

Project Setup (Each Project)

# Open project in Claude Code
cd ~/your-project && claude

# Initialize project
/setup-project

Update Commands

# Update Claude Copilot itself
cd ~/.claude/copilot && claude
/update-copilot

# Update project with latest
cd ~/your-project && claude
/update-project

Knowledge Repository (Optional)

# Create shared knowledge repository
/knowledge-copilot

Individual Components

Memory Copilot

cd ~/.claude/copilot/mcp-servers/copilot-memory
npm install
npm run build

Skills Copilot

cd ~/.claude/copilot/mcp-servers/skills-copilot
npm install
npm run build

Agents

# Copy all agents to project
cp ~/.claude/copilot/templates/agents/*.md ~/your-project/.claude/agents/

# Copy specific agent
cp ~/.claude/copilot/templates/agents/ta.md ~/your-project/.claude/agents/

Commands

# Copy all commands to project
cp ~/.claude/copilot/templates/commands/*.md ~/your-project/.claude/commands/

# Copy specific command
cp ~/.claude/copilot/templates/commands/protocol.md ~/your-project/.claude/commands/

Feature Cheat Sheet

Decision Matrix: "I want to..."

File Locations Reference

Project Level (~/your-project/)

your-project/
├── .mcp.json                    # MCP server configuration
├── CLAUDE.md                    # Project instructions (auto-loaded by Claude)
└── .claude/
    ├── commands/                # /protocol, /continue
    │   ├── protocol.md
    │   └── continue.md
    ├── agents/                  # 14 lean agents (under 120 lines each)
    │   ├── ta.md               # Tech Architect
    │   ├── me.md               # Engineer
    │   ├── qa.md               # QA Engineer
    │   ├── sec.md              # Security
    │   ├── doc.md              # Documentation
    │   ├── do.md               # DevOps
    │   ├── sd.md               # Service Designer
    │   ├── uxd.md              # UX Designer
    │   ├── uids.md             # UI Designer
    │   ├── uid.md              # UI Developer
    │   ├── cw.md               # Copywriter
    │   └── kc.md               # Knowledge Copilot
    └── skills/                  # Project-specific skills (optional)

User Level (~/.claude/)

~/.claude/
├── copilot/                     # Claude Copilot framework (cloned repo)
│   ├── mcp-servers/            # MCP server source code
│   │   ├── copilot-memory/     # Memory Copilot
│   │   └── skills-copilot/     # Skills Copilot
│   ├── templates/              # Source templates
│   │   ├── agents/             # Base agents
│   │   └── commands/           # Base commands
│   └── docs/                   # Documentation
├── knowledge/                   # Global knowledge repository (optional)
│   ├── knowledge-manifest.json
│   └── .claude/
│       └── extensions/         # Global agent extensions
└── memory/                      # Memory Copilot storage (auto-created)
    └── [workspace-hash]/       # Per-project database
        └── memory.db

Machine Level (~/.claude/copilot/)

~/.claude/copilot/              # Framework installation
├── .claude/
│   ├── commands/               # Machine-level commands
│   │   └── setup.md           # /setup command
│   └── agents/                # (Unused - agents are in templates/)
├── mcp-servers/               # MCP servers
│   ├── copilot-memory/
│   │   ├── src/
│   │   ├── package.json
│   │   └── build/             # Built JS files
│   └── skills-copilot/
│       ├── src/
│       ├── package.json
│       └── build/             # Built JS files
└── templates/                 # Source for project setup
    ├── agents/
    ├── commands/
    └── mcp.json

Memory Storage (~/.claude/memory/)

~/.claude/memory/
└── [workspace-id]/            # Unique per project (path hash)
    └── memory.db              # SQLite database
        ├── initiatives        # Current and archived initiatives
        ├── memories           # Decisions, lessons, context
        └── embeddings         # Vector search data

The Five Pillars

1. Memory Copilot

MCP server providing persistent memory across sessions

Environment:

• MEMORY_PATH: Base storage path (default: ~/.claude/memory)
• WORKSPACE_ID: Explicit workspace identifier (optional, defaults to path hash)

2. Agents

14 lean agents with on-demand skill loading

Agents are under 120 lines each and auto-load relevant skills via skill_evaluate(). Shared boilerplate is extracted to the "Agent Shared Behaviors" section in CLAUDE.md.

3. Skills Copilot

MCP server for on-demand skill loading and knowledge search

Environment:

• KNOWLEDGE_REPO_PATH: Project-specific knowledge (optional)
• ~/.claude/knowledge: Global knowledge (auto-detected)
• DATABASE_URL: PostgreSQL for private skills (optional)

4. Task Copilot

CLI tool (tc) for ephemeral PRD, task, and work product storage

Agents store detailed work products here instead of returning them to the main session, reducing context usage by ~96%.

Environment:

• TASK_DB_PATH: Database storage path (default: ~/.claude/tasks)
• WORKSPACE_ID: Links to Memory Copilot workspace (auto-detected)

5. Protocol

Commands enforcing battle-tested workflows

Common Workflows

First Time Setup

# 1. Clone framework
mkdir -p ~/.claude && cd ~/.claude
git clone https://github.com/Everyone-Needs-A-Copilot/claude-copilot.git copilot

# 2. Machine setup
cd ~/.claude/copilot && claude
# Then: Read @~/.claude/copilot/SETUP.md and set up Claude Copilot on this machine

# 3. Project setup
cd ~/your-project && claude
/setup-project

# 4. Start working
/protocol

Daily Work

# Resume previous session
/continue

# Start fresh task
/protocol

# At end of session
# Memory Copilot auto-saves via initiative_update

Adding Company Standards

# Create global knowledge repository
/knowledge-copilot

# Create extension
# Edit: ~/.claude/knowledge/.claude/extensions/ta.extension.md

# All projects now use your extended agents (auto-detected)

Updating

# Update Claude Copilot
cd ~/.claude/copilot && claude
/update-copilot

# Update each project
cd ~/your-project && claude
/update-project

Quick Troubleshooting

Verification Commands

# Check MCP servers built
ls ~/.claude/copilot/mcp-servers/*/build/

# Check project setup
ls .claude/agents/ .claude/commands/

# Check memory database
ls ~/.claude/memory/*/memory.db

# Check knowledge repo
ls ~/.claude/knowledge/knowledge-manifest.json

# Rebuild MCP servers
cd ~/.claude/copilot/mcp-servers/copilot-memory && npm install && npm run build
cd ~/.claude/copilot/mcp-servers/skills-copilot && npm install && npm run build

Extension System

Extension Types

Two-Tier Resolution

Minimal Global Knowledge Repo

mkdir -p ~/.claude/knowledge/.claude/extensions
cd ~/.claude/knowledge

# Create manifest
cat > knowledge-manifest.json << 'EOF'
{
  "version": "1.0",
  "name": "my-company",
  "description": "Company-specific agent extensions"
}
EOF

# Create extension (example)
cat > .claude/extensions/ta.extension.md << 'EOF'
---
extension_type: extension
target_agent: ta
---

## Company Architecture Standards

[Your company's architecture guidelines...]
EOF

Agent Routing

Agents automatically route to each other based on expertise:

Memory Copilot Session Management

Starting Work

# Fresh task
/protocol

# Resume previous work
/continue

During Work

Memory Copilot automatically tracks:

• Decisions made
• Lessons learned
• Progress updates
• Key files touched

Ending Work

Update initiative with:

initiative_update({
  completed: ["Tasks finished"],
  inProgress: "Current state",
  resumeInstructions: "Next steps for /continue",
  lessons: ["Insights gained"],
  decisions: ["Choices made"],
  keyFiles: ["file1.ts", "file2.ts"]
})

Configuration Quick Reference

.mcp.json (Project Level)

{
  "mcpServers": {
    "copilot-memory": {
      "command": "node",
      "args": ["/Users/you/.claude/copilot/mcp-servers/copilot-memory/build/index.js"],
      "env": {
        "MEMORY_PATH": "/Users/you/.claude/memory"
      }
    },
    "skills-copilot": {
      "command": "node",
      "args": ["/Users/you/.claude/copilot/mcp-servers/skills-copilot/build/index.js"],
      "env": {
        "KNOWLEDGE_REPO_PATH": "/path/to/project/knowledge"
      }
    }
  }
}

Environment Variables

Links to Full Documentation

Quick Tips

• Memory survives sessions: Use /continue to resume exactly where you left off
• Agents are stateless: They don't remember previous conversations (use Memory Copilot for that)
• Skills load on demand: No need to preload, Skills Copilot fetches when needed
• Knowledge is two-tier: Project knowledge overrides global knowledge
• Extensions auto-merge: Base agent + your extension = customized agent
• No time estimates: Framework uses phases, priorities, and complexity instead
• Git-friendly: All config files are plain text, commit .claude/ and CLAUDE.md
• Update regularly: Run /update-copilot and /update-project when framework updates

This Card: Print or bookmark for quick reference!

Quick Start: Run /protocol and let Claude Copilot guide you.