Pro Liberation] What is Claude Code? From the basics to the differences from Gemini and ChatGPT!

Among the many AI coding tools on the market, Claude Code, developed by Anthropic, has attracted particular attention.

Claude Code, previously available only on the Max plan, will be available on the “Claude Pro” plan in June 2025.

However, some may think, “I can code well enough with regular Claude, and I don’t feel the need to go to the trouble of using Claude Code …….

However, Claude Code is a tool that allows detailed development to areas that are inaccessible with regular Claude.

This article provides a thorough explanation of Claude Code based on the latest information on how it has been used and how it differs from other AI tools such as ChatGPT and Gemini.

What is Claude Code?

Claude Code is an AI coding tool developed by Anthropic that runs on the terminal.

Unlike common IDE code assistance, it has an “agent-type ” mechanism that understands the developer’s instructions and proceeds with the work on its own.

For example, if you ask in natural language to “add this feature,” Claude Code will automatically perform multiple steps to actually rewrite and execute the code.

A key feature of this tool is that it works as if you were a member of the development team.

technical basis

Claude Code is an AI coding tool based on Anthropic’s large-scale Claude 4 series of language models (Opus and Sonnet).

These models have high reasoning power and can understand long sentences, allowing them to handle even the most complex tasks.

MCP allows Claude Code to work securely with local files, IDEs, and even external apps such as Notion and GitHub.

While many AI tools help with code completion and modification within the IDE, Claude Code puts the terminal at the center of development and works as a tool that “executes” the developer’s instructions.

Main Functions and Capabilities

The functions that Claude Code has are not independent of each other, but work in conjunction with each other to function as an “autonomous development partner”.

• Understanding of the entire codebase: Understand the structure and dependencies of the entire project, and work with a wide range of codebases.
• Multi-File Editing: Complex changes across multiple files can be performed simultaneously while maintaining consistency.
• Terminal and Git integration: Automate commit creation and pull request generation.
• Autonomous error correction: reads error logs, analyzes root cause of problems, and automatically attempts to correct them.
• Safety: A tiered authorization system is in place to safely increase the level of automation while building trust with users.

Claude Code is now available on the Pro plan!

The terminal version of Claude Code will be available on the Pro plan around June 2024.

However, the tool eventually became one of the pillars of Anthropic’s premium services when it was officially incorporated as a benefit of the Claude Pro and Max plans, starting June 4, 2025.

The Pro plan allows you to use Claude Code via the terminal to the extent included in the monthly fee.

At present, however, it is intended for use in small to medium-sized repositories, and is mainly suited for bug fixing and simple scripting.

It is not yet recommended for use in large projects.

I actually used Claude Code.

You may be thinking, “I kind of get that it’s awesome, but what is Claude Code really like?” Some of you may be thinking, “I know it’s kind of awesome, but what is Claude Code really like?

For the first time, the author used Claude Code to create a tool to check whether text used in advertisements for cosmetics and quasi-drugs violates the “Pharmaceutical Affairs Law.

Once you get used to it, you will likely be able to complete tool development in a much shorter time.

The following are some of the points that impressed us when we actually used the system.

• Initial setup may seem a bit of a hurdle if you are not familiar with it
• Can generate a web application in about 10 minutes with instructions without requirements definition.
• Smooth acquisition of information from MCP integration with tools like Notion
• Easily manageable as it can be called from editors such as VS Code and Cursor
• Less likely to consume tokens if used after defining requirements in ChatGPT or Gemini

1. If you are not familiar with the system, the initial setup may seem a little difficult.

To use Claude Code, some technical preparation is required.

The basic setup consists of the following three steps, but other settings are required for actual use with VScode, Cursor, etc.

1. Installation of Node.js (version 16 or higher)
2. Install tools with npm
3. Create MCP server configuration file

Claude Code is provided as a CLI tool that runs in Node.js and works with Anthropic’s Claude 4 (Opus and Sonnet) in the background.

After installation, you must create a basic configuration file that specifies the address of the “Model Context Protocol (MCP)” server to connect Claude to your local development environment and external tools.

These settings allow Claude Code to work with terminals, editors, files, etc.

Therefore, there is a possibility that you may have a little difficulty with the initial setup that must be done first.

2. Generate a web application in about 10 minutes, even with instructions without requirements definition.

Claude Code’s greatest appeal is its unparalleled speed of prototyping. Claude Code’s greatest appeal is its

Even highly abstract, vague and rough instructions have the ability to be understood precisely and shaped into functional applications.

Claude Code generates working prototypes from ideas such as “this is what I want” without a clear design document or detailed specifications.

In fact, it took the author only about two hours to build the “drug law check application” from requirements definition to some state.

Even more remarkable is Claude Code’s ability to continue working autonomously.

Rakuten’s internal testing has successfully continued coding uninterrupted for up to seven hours after a single prompt for direction.

The results indicate that the AI is capable of continuing to perform tasks on a longer-term and sustained basis, rather than just providing short-term supplementary assistance.

3. Smooth acquisition of information from MCP linkage with tools such as Notion

One of the key features of Claude Code is its powerful integration capabilities with external tools through “MCP”.

MCP is a mechanism that allows Claude to exchange information with other applications in a secure and standardized manner, thereby seamlessly integrating Claude Code with commonly used tools such as Notion, Slack, and GitHub.

For example, the following processes can be automated by linking MCPs, all in a single series of tasks.

1. Claude Code reads the specifications written in Notion.
2. Automatic code generation based on information read
3. Push the completed code to a repository on GitHub and create a pull request

In addition, linkages can be added to these processes, such as posting progress and completion reports to Slack.

By reducing the amount of manual copy-pasting and checking, developers can focus on more creative tasks, and overall team productivity will increase dramatically.

4 . Easy to manage as it can be called from editors such as VS Code and Cursor

Claude Code is “terminal-centric” in design, allowing you to receive AI suggestions simply by invoking the CLI from within the editor.

Developers who are terminalists, as well as developers who normally work with VS Code or Cursor, can take advantage of Claude Code without leaving their familiar environment.

Specifically, the code changes and modifications suggested by Claude Code can be displayed directly within the IDE.

Claude Code’s can make suggestions that take into account the context of the entire project, not just a piece of code.

By being available in the editor tool, Claude Code is more than a “helper”; it is a partner in the project, providing advanced understanding and recommendations.

In addition, Claude Code can be used locally to load and summarize document files stored on your own computer.

Claude Code breaks through the limitations felt when using AI in the browser and improves the convenience of local development.

5 . Less likely to consume tokens if used after defining requirements with ChatGPT or Gemini.

Claude Code has a token limit, even for the Pro and Max plans, which can be reached quickly if there are many exchanges.

For example, ChatGPT and Gemini are used to organize thoughts such as “what functions to create” and “what data structures to use.

The summarized requirements definition can then be passed to Claude Code, which can generate highly accurate code with minimal interaction.

# 薬機法リスクチェッカー 要件定義書

---

## 1. 概要

### 1.1. 本書の目的
本書は、Webアプリケーション「薬機法リスクチェッカー」の開発にあたり、その仕様、機能、要件を明確に定義することを目的とする。開発者および関係者は、本書を正として開発を進める。

### 1.2. 開発背景と目的
美容業界（サロン、化粧品メーカー）の広告・制作担当者や、専門知識を持たない外部デザイナー・ライターは、広告表現における薬機法遵守に課題を抱えている。自身の知識への不安や、専門部署への確認で生じる頻繁な手戻りにより、業務効率が低下している。

本アプリは、これらの担当者が広告文のラフ案を手軽に一次チェックできる環境を提供することで、**手戻りを削減し、制作スピードを向上させること**を主目的とする。

---

## 2. アプリケーションの基本仕様

### 2.1. アプリ名（案）
薬機法リスクチェッカー

### 2.2. コンセプト
薬機法に不慣れな広告・制作担当者のための、手軽で直感的な一次チェック＆リライト支援ツール

### 2.3. ターゲットユーザー
- 美容サロンの広告担当者、オーナー
- 化粧品メーカーの広告担当者
- 美容系の記事・LP・バナー等の制作を請け負うデザイナー、ライター

---

## 3. 機能要件

### 3.1. 画面構成
本アプリは、以下の2つの主要なタブで構成される。

1.  **リスクチェッカータブ:** メイン機能。テキストを入力し、薬機法リスクをチェックする。
2.  **薬機法簡単ガイドタブ:** サブ機能。薬機法初心者向けの解説コンテンツを提供する。

### 3.2. リスクチェッカー機能
#### 3.2.1. 入力仕様
1.  ユーザーは、まずチェックしたい文章の種類を以下の選択肢から選ぶ。
    - `キャッチコピー`
    - `商品説明文`
    - `お客様の声`
2.  指定されたテキストエリアに、チェックしたい文章をコピー＆ペーストで入力する。
3.  「チェック開始」ボタンをクリックすることで、チェック処理が実行される。

#### 3.2.2. 出力・表示仕様
「チェック開始」ボタン押下後、結果表示エリアに以下の情報が整形されて表示される。

1.  **総合リスクレベル:** テキスト全体のリスクを「高」「中」「低」の3段階で表示する。
    > 例: `総合リスクレベル: 高`
2.  **リスク件数サマリー:** 検出されたリスクの件数を表示する。
    > 例:
    > - 総検出数: 3
    > - 高リスク: 1
    > - 中リスク: 2
    > - 低リスク: 0
3.  **詳細結果（2カラム表示）:**
    - **左カラム（入力文表示エリア）:**
        - ユーザーが入力した文章がそのまま表示される。
        - AIによって指摘された問題箇所がハイライト（例: `<mark>`タグによる黄色マーカー）される。
    - **右カラム（解説エリア）:**
        - 指摘事項がリスト形式で表示される。
        - 各指摘事項には、以下の情報が含まれる。
            - **問題の表現:** ハイライトされた具体的なテキスト
            - **抵触の理由:** なぜ薬機法に抵触する可能性があるかの具体的解説
            - **リスクレベル:** 「高」「中」「低」の個別評価
            - **代替表現の提案:** 3つ程度の具体的な言い換え候補
4.  **全文リライト案:**
    - 指摘事項を反映し、広告効果を維持しつつ薬機法に準拠するようAIがリライトした全文を提示する。
5.  **インタラクション:**
    - 左カラムのハイライト箇所をクリックすると、右カラムの対応する指摘事項解説へ自動でスクロールする。

### 3.3. 薬機法簡単ガイド機能
- **目的:** 薬機法の知識がない、または浅いユーザーへの啓蒙と、アプリへの理解促進。
- **コンテンツ:** 化粧品広告等で頻出するNG表現や、その考え方、OKな言い換え例などを、図や平易な文章を用いて分かりやすく解説する。

---

## 4. AIおよびシステム要件

### 4.1. AIチェックロジック
- **知識ソース:**
    1.  **プライマリ:** Notionに格納された、薬機法に関するドキュメント（NoteBookLM, Gemini等で精査・整理された情報）。
    2.  **補助:** アプリケーションのディレクトリ内に配置された「NG広告例.csv」ファイル。
- **連携方法:** Claude Code APIは、MCP（iPaaS連携ツール等を想定）を介してNotionから情報を取得する。
- **出力形式:** AIは、フロントエンドでの処理を容易にするため、必ず以下のJSON形式で応答を返すこと。

```json
{
  "overall_risk": "高" | "中" | "低",
  "risk_counts": {
    "total": Number,
    "high": Number,
    "medium": Number,
    "low": Number
  },
  "issues": [
    {
      "fragment": "問題のある表現",
      "reason": "抵触する理由...",
      "risk_level": "高" | "中" | "低",
      "suggestions": ["代替案1", "代替案2", "代替案3"]
    }
  ],
  "rewritten_text": "リライトされた全文..."
}

Thus, by dividing the role of each AI, both speed and accuracy of work can be achieved while saving tokens.

This approach is especially effective in environments with limited usage, such as the Pro plan.

What is different about Claude Code compared to other AI tools?

With the growing selection of AI coding tools available, the question is naturally, “What makes Claude Code different from other tools?” is a natural question.

Compare the features of each tool in terms of programming accuracy, user-friendliness, cost-effectiveness, and relationship to existing tools.

Claude Code leads by far in programming accuracy

Claude Code is far ahead of other AI tools in terms of accuracy in practical programming tasks.

Especially noteworthy are the results of SWE-bench Verified, an evaluation index based on actual issues such as bugs and functional improvements on GitHub.

The results show that Claude Code has the power to provide solutions to complex problems that can occur in real-world development on a practical level.

It is characterized by its strength in elements required in practice, such as understanding specifications, understanding context, and dealing with multi-file configurations.

Ease of use and data analysis for users is robust with ChatGPT

The strength of ChatGPT lies in its unparalleled versatility and user-friendly interface.

Especially for novice programmers and data scientists who want to perform analysis in an environment such as Jupyter Notebook, ChatGPT is very powerful because it can easily generate and modify code in an interactive format.

With extensive library knowledge and excellent troubleshooting capabilities, it offers stable performance in a wide range of applications from script generation to data analysis and visualization.

Gemini is superior in terms of cost performance and data volume

Gemini sets itself apart from other AIs in terms of cost-effectiveness and the amount of data it can process.

The main attraction is the ability to process large amounts of data at low cost, backed by Google’s powerful infrastructure.

Therefore, Gemini is a very compatible tool for needs such as “I want it to be inexpensive” or “I want it to read and respond to long specifications.

Claude Pro Fee Structure

Claude’s pricing plans range from free versions to customized plans for corporate clients.

Featured Claude Code is available on the $20/month “Pro” plan and above.

plan	Monthly Fee (USD)	Japanese Yen equivalent (approximate)	Claude Code support	Key Features
Free	$0	free	✕	Basic chat functionality (limited use)
Pro	$20	Approx. 3,300 yen	period	Approximately 45 messages per 5 hours, priority access to the latest models
Team	$30	Approx. 4,950 yen	✕	Pro plan features plus team use features
Max.	From $100	–	period	Significantly more usage than the Pro plan, highest level of security

Claude Code Application Examples

Claude Code is more than just a code generator; it has the potential to revolutionize the entire development workflow.

Examples of its use range from rapid prototyping, to secure development in a local environment, to even helping teams learn.

Let’s look at how it can be utilized.

Generate prototypes with fast coding anyway

Claude Code is highly effective in “prototyping” situations, where ideas are quickly shaped and tested.

Furthermore, Claude Code has proven itself not only in prototyping, but also in specialized and advanced development tasks such as converting data analysis code into a production data pipeline.

Claude Code will increasingly become more than just a code generation tool, but a “development booster” to accelerate the speed of innovation.

Improve AI development capabilities in local environments

One of Claude Code’s greatest strengths is, The ability to work directly in the local environment through the MCP through the MCP.

With MCP, you can take advantage of AI assistance even for files on the local level.

There is no need for complicated file sharing or data importing.

Claude Code directly accesses the local file system to read and modify code, generate files, etc., allowing users to receive AI assistance in their usual development environment.

Claude Code can be said to be an AI tool that has established a very flexible and practical position as an AI tool that can provide full-fledged development support that goes deep into the local area without having to rely on the cloud.

Version control of generated code and apps on GitHub

Claude Code is also very compatible with standard development workflows.

This allows AI-generated artifacts to be seamlessly integrated into version control systems in exactly the same process as human work.

It is expected to facilitate collaboration between AI and humans and increase the transparency and manageability of the entire development process.

Operating as a lecturer to catch up on new information

Claude Code is not only a development support tool, but also an educational support tool.

AI’s unique support is especially effective in helping new team members get up and running.

For example, in the field of software development, understanding ” legacy code ” that has been used for many years is often a major obstacle, and until now, much time and effort has been devoted by veteran engineers to explaining and taking over such code.

Claude Code can read its entire code base, understand the context, and immediately provide specific and detailed explanations to new members’ questions, greatly reducing training costs.

In addition, AI-based training is always of consistent quality, which has the added benefit of reducing human variability in teaching methods.

Claude Code can be used not only to streamline development work, but also as a tool that contributes to improving the skills of the entire team and the efficiency of handing over work, establishing a “build while you grow” style that integrates education and practice.

Frequently asked questions about Claude Code and Claude Pro

Finally, we answer some frequently asked questions about Claude Code and Claude Pro.

Can Claude Code be used by beginners?

Claude Code is a tool that can be used by beginners, but some may struggle with the initial setup.

There are some technical hurdles to the initial setup, such as knowledge of Node.js and editing configuration files.

For example, if you are on Windows, you cannot use Claude Code without installing Node.js or Ubuntu unless you activate Linux on the terminal.

However, once set up, Claude Code does not require in-depth technical knowledge to use, as the core operation is performed via natural language instructions.

However, it is important to remember that a basic understanding of programming is essential to evaluate the quality of the generated code and to modify and guide it appropriately.

Is Claude Code available in a free version?

Access to Claude Code is a premium feature and requires a monthly subscription to the 20 Claude Pro plan or higher.

Advanced coding assistance features require an upgrade to a paid plan.

Note that Claude Code is available even in the free version when using the API.

Does Claude Pro support Japanese?

Claude in general is able to handle Japanese, and the underlying model has very high Japanese processing capability.

Claude’s ability to understand the detailed nuances of the Japanese cultural background and context, rather than simply translating, is another reason why he is so highly regarded.

Can Claude Pro be used on multiple devices?

A single Claude Pro account can be accessed simultaneously from multiple devices, including a web browser, an iOS/Android mobile app, and a desktop terminal running Claude Code.

summary

In this article, we explained Claude Code from many angles, from its basic concepts to its use in the Pro plan, and how it differs from major competitors such as ChatGPT and Gemini.

Claude Code is not just a code completion tool, but an “agent” that runs autonomously on the terminal and performs particularly well in solving complex real-world tasks.

Available with the $20/month Pro plan, Claude Code will revolutionize your development workflow.

On the other hand, each tool has its own strengths, such as ChatGPT for ease of use and Gemini for cost and data volume.

Use the information in this article to establish the development style that best suits your own development style and objectives.