Claude Copilot Quick Start (15 Minutes)

Claude Copilot Quick Start (15 Minutes)

This guide gets you up and running with Memory Copilot only - the fastest path to persistent session memory.

What You Get

With this minimal setup:

This is perfect if:

• You want to try Memory Copilot first
• You need session persistence but not agents
• You want the fastest possible setup
• You plan to upgrade to the full framework later

You'll miss out on:

• 14 lean agents with auto-skill loading (Tech Architect, QA, Security, etc.)
• Context-aware skill evaluation (skill_evaluate)
• Knowledge repository integration
• Agent-First Protocol workflow

Setup (5 Steps)

Step 1: Clone the Repository

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

Step 2: Build Memory Copilot

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

Verify: Check that dist/index.js exists:

ls ~/.claude/copilot/mcp-servers/copilot-memory/dist/index.js

Step 3: Create Memory Directory

mkdir -p ~/.claude/memory

Step 4: Configure Your Project

In your project directory, create .mcp.json:

{
  "mcpServers": {
    "copilot-memory": {
      "command": "node",
      "args": ["/Users/YOUR_USERNAME/.claude/copilot/mcp-servers/copilot-memory/dist/index.js"],
      "env": {
        "MEMORY_PATH": "/Users/YOUR_USERNAME/.claude/memory",
        "WORKSPACE_ID": "my-project"
      }
    }
  }
}

IMPORTANT: Replace:

• /Users/YOUR_USERNAME with your actual home directory path
• my-project with your project name (used as the memory workspace identifier)

Step 5: Create Continue Command

Create .claude/commands/continue.md in your project:

mkdir -p .claude/commands

Then download or copy the continue command from:

cp ~/.claude/copilot/.claude/commands/continue.md .claude/commands/

Verification

1. Restart Claude Code

Close and reopen Claude Code in your project directory.

2. Check MCP Connection

Run /mcp in Claude Code. You should see:

Connected MCP Servers:
● copilot-memory

3. Test Memory Tools

Try these commands in Claude:

# Start a new initiative
initiative_start({
  title: "Test Memory Setup",
  description: "Verifying Memory Copilot works"
})

# Store a memory
memory_store({
  type: "decision",
  content: "Memory Copilot is working!",
  tags: ["setup", "test"]
})

# Search memories
memory_search({ query: "working" })

# Get current initiative
initiative_get()

If these work, you're all set!

4. Test /continue Command

Say: /continue

You should see Claude load your current initiative and offer to resume work.

Daily Usage

Starting Work

/continue

Claude will:

1. Load your current initiative
2. Show recent progress
3. Review key decisions
4. Present next steps
5. Ask what you want to work on

During Work

Memory is automatically saved when you:

• Make decisions
• Complete tasks
• Learn lessons
• Touch important files

Ending Work

Just close Claude Code. Your progress is automatically saved.

To explicitly update before ending:

initiative_update({
  completed: ["Task 1", "Task 2"],
  inProgress: "Currently working on X",
  resumeInstructions: "Next: Start with Y",
  lessons: ["Learned that Z works better"],
  decisions: ["Decided to use approach A"],
  keyFiles: ["src/main.ts", "src/utils.ts"]
})

Troubleshooting

"MCP server not found"

Check paths in .mcp.json:

• Paths must be absolute (no ~ or $HOME)
• Verify file exists: ls /path/to/dist/index.js
• Match your actual home directory path

"Build failed"

Install build tools:

macOS:

xcode-select --install

Linux:

sudo apt-get install build-essential python3

Windows:

npm install --global windows-build-tools

Then rebuild:

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

"Memory not persisting"

Check memory directory:

ls ~/.claude/memory/

Should show a subdirectory with your workspace ID.

Check workspace ID:

• Make sure WORKSPACE_ID in .mcp.json matches across sessions
• Don't change the workspace ID or you'll lose access to previous memories

"Commands not working"

Verify command file:

cat .claude/commands/continue.md

Should show the continue command content.

Restart Claude Code to reload commands.

Upgrade Path

When you're ready for the full framework:

Option 1: Full Setup Command

cd ~/.claude/copilot
# Open in Claude Code
/setup-copilot

Then in your project:

/setup-copilot

This will add:

• All 14 lean agents
• Skills Copilot (with skill_evaluate)
• Task management via tc CLI
• /protocol command
• Full command set

Option 2: Manual Addition

Keep your .mcp.json and add components one at a time:

Add Skills Copilot:

1. Build: cd ~/.claude/copilot/mcp-servers/skills-copilot && npm install && npm run build
2. Add to .mcp.json (see templates/mcp.json)

Add Agents:

cp ~/.claude/copilot/.claude/agents/*.md .claude/agents/

Add Commands:

cp ~/.claude/copilot/.claude/commands/protocol.md .claude/commands/

Add Task Management: Install the tc CLI tool for PRD, task, and work product management.

What's Next?

With Memory Copilot running, you can:

1. Use /continue daily - Start every session by resuming context
2. Store decisions - Call memory_store for important choices
3. Search history - Use memory_search to find past context
4. Track initiatives - Manage work with initiative_* tools
5. Upgrade when ready - Add agents and skills later

Need Help?

• Full Setup Guide: ~/.claude/copilot/SETUP.md
• Complete Documentation: ~/.claude/copilot/README.md
• GitHub Issues: https://github.com/Everyone-Needs-A-Copilot/claude-copilot/issues

Summary

You now have:

• Persistent memory across sessions
• Context preservation between work sessions
• Decision and lesson tracking
• /continue command for seamless resumption

This is the foundation. Add agents and skills when you need specialized expertise.

Next session: Just say /continue and pick up where you left off.