Advanced Claude Techniques

Once you’re comfortable with basics, try these features for even better Claude utilization.

CLAUDE.md (Project Settings File)

Creating a CLAUDE.md file at your project root lets Claude automatically understand your project’s rules and context.

Example CLAUDE.md contents:

# Project Overview
 
This project is an internal expense management system.
 
## Tech Stack
- Frontend: React + TypeScript
- Backend: Python (FastAPI)
- Database: PostgreSQL
 
## Coding Standards
- Add Japanese comments explaining variable names
- Display error messages in Japanese
- Always write tests
 
## Notes
- Don't handle production data
- Store API keys in .env file

Benefits:

• No need to explain “this project is…” every time
• Generates code following project rules
• Entire team uses Claude with same settings

Location and Usage:

Global settings apply to all projects. If project-specific settings exist, both are loaded.

MCP Servers (External Service Integration)

Using MCP (Model Context Protocol), Claude can integrate with external services.

What is MCP?

A system to extend Claude’s capabilities. For example:

• Read and write Slack messages
• Get Shopify product data
• Run database queries directly
• Manipulate the file system

Setup

Add MCP servers in VS Code settings (settings.json).

How to open settings file:

1. Open command palette with Cmd + Shift + P
2. Select “Preferences: Open User Settings (JSON)”

Example configuration (Slack MCP):

{
  "claude.mcpServers": {
    "slack": {
      "command": "npx",
      "args": ["-y", "@anthropic/mcp-server-slack"],
      "env": {
        "SLACK_BOT_TOKEN": "xoxb-your-token-here",
        "SLACK_TEAM_ID": "T0123456789"
      }
    }
  }
}

Types of MCP Servers

MCP servers can be categorized into three types based on what they provide:

Example: Shopify-related MCPs

Common MCP Servers

Information/Docs Providers (reference, often no auth required)

Data Providers (real data access, auth required)

Tool/Action Providers (execute operations)

For more, see MCP Official Repository  or Awesome MCP Servers .

Shopify Dev MCP

A helpful MCP server for Shopify app development. Runs locally without authentication.

Type: Information/Docs Provider (no access to real data)

Features:

• Search Shopify API documentation
• Explore Admin GraphQL API schema
• Get info on Polaris, Hydrogen, Storefront Web Components
• Shopify Functions development support

Example configuration:

{
  "claude.mcpServers": {
    "shopify-dev": {
      "command": "npx",
      "args": ["-y", "@shopify/dev-mcp"]
    }
  }
}

Shopify MCP (Data Provider)

An MCP server to access real data in your Shopify store. Requires authentication.

Type: Data Provider (access to actual products, inventory, orders)

Use cases:

For EC site operations:

List all products with inventory below 10

Show the top 10 selling products from last week

Update the price of "Summer T-shirt" to 2,980 yen

Slack Integration

With the Slack MCP server configured, Claude can directly interact with Slack.

What you can do

Information gathering:

Summarize today's messages in the #general channel

Check unread mentions from @Tanaka

Creating posts:

Post a weekly progress report to #dev-team with:
- Feature A implementation complete
- Feature B is in testing

Analysis:

Analyze the past week of inquiries in #support channel
and summarize the most common questions

What you need

1. Slack Bot Token: Created in your Slack workspace
2. Team ID: Found in Slack settings
3. Required permissions: channels:read, chat:write, etc.

Ask your mentor for detailed setup instructions.

Sub-Agents (Automatic Task Splitting)

Claude automatically splits complex tasks into smaller tasks and processes them in parallel. These are called sub-agents.

When are they used?

• When searching/editing multiple files simultaneously
• When performing large refactoring
• When running tests in parallel

Example

In all JavaScript files in the project,
replace var with let or const

Claude automatically launches:

1. An agent to find target files
2. Agents to modify each file
3. An agent to verify changes

and processes them efficiently.

How to check

When sub-agents are running, you may see “Agent” displayed in the chat. Since multiple processes run simultaneously, it may take longer to get results than for simple tasks.

Agent Teams

Agent Teams is an advanced feature that builds on sub-agents. Users can create teams of agents with defined roles that work together collaboratively.

Difference from Sub-Agents

Example Roles

With Agent Teams, you can define specialists such as:

• QA Expert: Test creation, quality checking
• Frontend Lead: UI/UX, React/Vue implementation
• Backend Lead: API, database, server processing
• Security Lead: Vulnerability checks, security reviews
• Documentation Lead: README, API specification writing

Usage Example

Join this project as a QA team and
improve the test coverage

From a QA expert perspective, suggests testing strategies and creates test code.

Current Status

Agent Teams is provided as an experimental feature. Basic sub-agent functionality is used automatically, but detailed team definition features will continue to evolve in the future.

Start by utilizing the automatic sub-agent splitting feature, and consider Agent Teams configuration as needed.