What is Claude Code as a Daily Driver: Claude.md, Skills, Subagents, Plugins, and MCPs? A Practical Overview
Claude Code is an AI coding agent that runs in your terminal. This post explains how to use it as a daily driver with Claude.md, Skills, Subagents, Plugins, and MCPs, including concrete examples and honest tradeoffs.
Claude Code is an AI-powered coding agent that operates directly in your terminal. It can read and write files, run commands, and interact with your codebase. The trend around "Claude Code as a daily driver" focuses on extending its capabilities with four key components: Claude.md, Skills, Subagents, Plugins, and MCPs (Model Context Protocol). This post explains what each of these is, how they fit together, and whether it makes sense to rely on Claude Code for your daily work.
What is Claude Code?
Claude Code is a command-line tool developed by Anthropic. It uses Claude 3.5 Sonnet (and newer models) to understand natural language instructions and execute them against your project. You install it via npm:
npm install -g @anthropic-ai/claude-code
Then run it in your project directory:
claude
This opens an interactive session where you can ask Claude to write code, refactor, debug, or run tests. It has access to your file system and can execute shell commands.
Claude.md: Your Project's Context File
Claude.md is a markdown file placed at the root of your project. It acts as a persistent context document that Claude reads at the start of every session. Think of it as a README for the AI. It tells Claude about your project's architecture, coding conventions, dependencies, and any constraints.
Example Claude.md:
# Project: MyApp
// stay current
AI & ML insights, weekly
Practical deep-dives on LLMs, developer tools, and AI engineering. No filler. Unsubscribe any time.
// written byFIG. AUTH-01
522
Mahmudul Haque Qudrati
CEO & ML Engineer
CEO and ML Engineer at Pristren. Builds AI-powered software for teams and writes about machine learning, LLMs, developer tools, and practical AI applications.
Without Claude.md, Claude starts each session with zero context. With it, you get consistent behavior across sessions. The file is checked into version control, so the whole team benefits.
## Skills: Reusable Capabilities
Skills are modular, reusable pieces of functionality that Claude Code can load. They are essentially plugins for Claude's behavior. A skill might be a custom prompt, a set of instructions, or even a small script that Claude can invoke.
Skills live in a `skills/` directory in your project or globally. For example, a "code-review" skill:
```markdown
# skills/code-review.md
You are a senior code reviewer. For each file:
1. Check for security vulnerabilities (SQL injection, XSS, etc.)
2. Verify error handling
3. Suggest performance improvements
4. Ensure test coverage
Output a numbered list of findings with severity (low/medium/high).
You can activate a skill by referencing it in your conversation or via a command. Skills are composable: you can chain multiple skills together.
Subagents: Delegating Tasks
Subagents are separate Claude Code instances that you can spawn to work on specific tasks in parallel. This is useful for large refactors or multi-file changes. You can delegate a subagent to, say, "refactor all API routes to use dependency injection" while you continue working on another part of the codebase.
Subagents share the same context (Claude.md and skills) but operate independently. They report back with a summary of changes. This is still experimental and can lead to conflicts if two subagents modify the same file.
Plugins: Extending the Toolchain
Plugins are external integrations that Claude Code can use. They are similar to Skills but are typically more complex and may involve external APIs or services. For example, a plugin could connect Claude to your Jira board, Slack, or a database.
Plugins are defined in a plugins/ directory and can be written in Python or JavaScript. A simple plugin might look like:
// plugins/weather.js
module.exports = {
name: 'weather',
description: 'Get current weather for a city',
async execute(args) {
const response = await fetch(`https://api.weather.com/${args.city}`);
return response.json();
}
};
Claude can then call this plugin when asked "What's the weather in Tokyo?"
MCPs: Model Context Protocol
MCP (Model Context Protocol) is a standardized way for Claude Code to interact with external tools and services. It's like a universal adapter. Instead of writing custom plugins for every service, you can use MCP servers that expose tools via a common interface.
For example, an MCP server for GitHub could provide tools like create_pr, list_issues, merge_branch. Claude Code can discover and use these tools automatically if the MCP server is running.
MCP is still evolving, but it promises to make Claude Code extensible without writing custom code for each integration.
Putting It All Together: A Daily Workflow
Here's how a typical daily session might look:
Start Claude Code in your project directory.
Claude reads Claude.md and loads your default skills.
You ask: "Add a new endpoint for user registration with email verification."
Claude uses its knowledge of your stack (from Claude.md) and follows your coding conventions.
It creates the necessary files, updates routes, and writes tests.
You review the changes, maybe ask for tweaks.
For a larger task, you spawn a subagent to handle the email template while you continue.
You use an MCP server to create a PR on GitHub directly from the terminal.
Costs and Limitations
Claude Code is not free. It uses API tokens that cost money. A typical session might consume $0.50-$2.00 in tokens, depending on complexity. Heavy daily use can add up to hundreds of dollars per month.
Other limitations:
Claude Code can make mistakes, especially with complex logic or edge cases.
It has a context window limit (currently 200k tokens). Very large codebases may need careful context management.
Subagents can conflict if not coordinated.
MCP servers add latency and potential failure points.
Is It Worth It?
For solo developers or small teams working on medium-sized projects, Claude Code with Claude.md, Skills, and a few plugins can significantly boost productivity. The key is to invest time in crafting a good Claude.md and reusable skills. Without that, the experience is mediocre.
For large teams or complex codebases, the cost and coordination overhead may outweigh the benefits. Start with a trial project and measure the time saved versus token cost.
Getting Started
Install Claude Code: npm install -g @anthropic-ai/claude-code
Create a Claude.md in your project root.
Add a skills/ directory with one or two skills.
Run claude and ask it to do something simple.
Iterate on your Claude.md based on what Claude gets wrong.
Ready to try Claude Code for your projects? Sign up for Zlyqor to manage your AI agent workflows: https://app.zlyqor.com/signup
Frequently Asked Questions
What is Claude Code?
Claude Code is a terminal-based AI coding agent by Anthropic. It can read and write files, run commands, and execute tasks based on natural language instructions. It uses Claude models and is installed via npm.
What is Claude.md?
Claude.md is a markdown file placed in your project root that provides persistent context to Claude Code. It describes your tech stack, conventions, and constraints, ensuring consistent behavior across sessions.
What are Skills in Claude Code?
Skills are reusable modules that define specific behaviors or tasks for Claude Code. They live in a skills/ directory and can be activated on demand. Examples include code review prompts or deployment scripts.
What are Subagents?
Subagents are separate Claude Code instances that work on delegated tasks in parallel. They share context but operate independently. Useful for large refactors but can cause file conflicts.
What are MCPs in Claude Code?
MCP (Model Context Protocol) is a standard for connecting Claude Code to external tools and services via a common interface. MCP servers expose tools like create_pr or list_issues that Claude can call.
How much does Claude Code cost?
Claude Code uses API tokens. Costs vary by usage but typically range from $0.50 to $2.00 per session. Heavy daily use can cost hundreds of dollars per month.
Is Claude Code suitable for large teams?
It depends. For small teams with well-defined projects, it can boost productivity. For large teams, costs and coordination overhead may be significant. Start with a trial project to evaluate.
