Claude Code is Anthropic's AI coding tool that you install and run in the terminal. As of October 1, 2026, it is not included in the free plan, and you can use it from Pro ($20 per month). Usage is shared with Claude chat and is counted by 5-hour sessions and weekly limits. Installation takes a single command.
We checked installation and first run in the Claude Code quickstart docs, pricing on the Claude pricing page, and cost management in the Claude Code cost docs. You do not need to be a developer to use it for reading and editing files in a folder.
Read this first in three lines
- Install: paste the one-line command for your operating system into the terminal, then run
claudeand log in. - Price: there is no free access. Pro is $20 per month and Max starts at $100 per month. When you use up your subscription limit, you can move to usage credits or API billing.
- Warning: if
ANTHROPIC_API_KEYis set on your computer, you are billed at API rates instead of using your subscription.
Installing Claude Code takes one command
Claude Code runs in the terminal (the window where you type commands). On a Mac, open the Terminal app. On Windows, open PowerShell. The installation methods the official docs recommend are in the table below.
| Environment | Install command |
|---|---|
| Mac, Linux, WSL | curl -fsSL https://claude.ai/install.sh | bash |
| Windows PowerShell | irm https://claude.ai/install.ps1 | iex |
| Windows CMD | curl -fsSL https://claude.ai/install.cmd -o install.cmd && install.cmd && del install.cmd |
| Homebrew (Mac) | brew install --cask claude-code |
| WinGet (Windows) | winget install Anthropic.ClaudeCode |

When the install finishes, open a new terminal and type claude --version. If a version number appears, it worked. If you see a message that claude cannot be found, the install location is not yet in your PATH (the list of folders where your system looks for commands). The official docs have a troubleshooting page for installation with the fix.
The requirements are modest. You need macOS 13 or later on a Mac, Windows 10 1809 or later on Windows, and Ubuntu 20.04 or later. You need at least 4GB of memory. On Windows, it works through PowerShell even without Git for Windows. South Korea is on Anthropic's list of supported countries.
If the terminal feels unfamiliar, there is a desktop app. The official docs describe a desktop app that lets you use Claude Code without a terminal. You can also use it on the web at claude.ai/code.
Log in and ask your first question
Go to the folder you want to work in and type claude. On first run, a browser opens with a login screen. Four kinds of accounts can log in.
- A Claude Pro, Max, Team, or Enterprise subscription account
- A Claude Console account (a developer account that pays for the API with prepaid credits)
- A cloud route such as Amazon Bedrock, Google Cloud, or Microsoft Foundry
- A company-run gateway
The official docs state that a free claude.ai account cannot use Claude Code. To switch accounts, type /login inside a session.
After you log in, ask your first question. The official docs use an example that asks about the folder contents, such as "what does this project do?" You can ask in Korean too. Claude Code reads the files it needs on its own. You do not have to paste files in each time.
Only a few commands come up often.
claude -c: continue the last conversation in this folder.claude -p "question": answer once and exit./clear: clear the conversation and start fresh./help: show the commands you can use.
Plans split into subscription and API
There are two main ways to use Claude Code. If you log in with a subscription, you use it within a flat monthly fee. If you log in with an API key or a Console account, you pay for the tokens you use. Tokens are the units of text the AI reads and writes.

We checked subscription prices on the pricing page on October 1, 2026.
| Plan | Monthly price (USD) | Claude Code | Usage |
|---|---|---|---|
| Free | 0 | Not included | Basic level |
| Pro | 20 (annual billing: 17 per month, 200 per year) | Included | 5x Free (per session) |
| Max | From 100 | Included | 5x or 20x Pro |
| Team Standard seat | 25 (annual billing: 20) | Included | Per-seat limits |
| Team Premium seat | 125 (annual billing: 100) | Included | 5x Standard |
Listed prices exclude tax. We compare every plan and the free limits in a separate post, Claude plan price comparison.
The API route has no monthly fee. According to the official cost docs, in enterprise deployments a developer spent an average of $13 per active day, or $150 to $250 per month. 90% of users stayed under $30 per day. These are averages for developer work. MCPY thinks a subscription is easier to budget for a marketer who uses it only occasionally.
Usage limits are counted together with chat
On a subscription, Claude Code and Claude chat share the same usage. The guide to using Claude Code with Pro and Max plans says so. If you chat a lot during the day, you have less left for code work.

Read limit messages as one of two kinds. "session limit" means you used up the 5-hour window, and "weekly limit" means you used up the weekly window. The official cost docs explain that these limits apply to all models, so switching models with /model does not lift them. However, if you see a message for a specific model family, such as "Opus limit", you can switch to a model from another family and keep working.
You have three options when you hit a limit.
- Wait for the reset. The message shows the reset time. On v2.1.234 and later, you can have interrupted work resume automatically after the reset.
- Turn on usage credits. Type
/usage-creditsto open the settings screen. You keep working past the limit, but extra charges apply. - Upgrade to a higher plan. Pro users who hit limits often should look at Max.
To see how much you have used, run /usage. Subscribers see a usage bar and the share of usage by skill and plugin. The dollar figure on the same screen is an estimate for API users. It is not a charge for subscribers.
Compare it with competing tools at the same price
Claude Code is not the only coding agent. Here are the tools you can pick in the same $20 range, based on official pages.

| Tool | Entry price (per month, USD) | Higher price (per month, USD) | Usage basis |
|---|---|---|---|
| Claude Code | 20 (Pro) | From 100 (Max) | 5-hour session, weekly limit |
| Codex | 20 (ChatGPT Plus) | 100, 200, 500 (Pro) | Plus gives an estimate of messages per 5 hours |
| Gemini CLI | 0 (free with a Google account) | Google AI Pro and Ultra | Free: 1,000 requests per day, Pro: 1,500 requests per day, Ultra: 2,000 requests per day |
The sources are the OpenAI Codex pricing docs and the Gemini CLI quota docs. The reference date is October 1, 2026. Gemini counts by number of requests and Claude counts by usage percentage, so you cannot compare limit sizes directly as numbers. MCPY did not rank the limit sizes in this table. For the ChatGPT plans, see our post on ChatGPT Pro $200 and $500 limits (in Korean).
Tasks marketers can try with Claude Code
The strength of Claude Code is that it reads and edits files directly. You do not have to copy and paste content as you do in chat. The examples below adapt the natural-language request style from the official docs to marketing work. They are not results MCPY measured itself.
- Tidy ad reports: put the downloaded CSV files in a folder and ask, "Combine impressions, clicks, and cost by campaign into a table."
- Check UTM links: ask it to find rows in a link list file where source and medium spellings are inconsistent and make them uniform.
- Bulk-edit content: have it check dozens of blog HTML files at once for title length or missing image descriptions.
The official docs recommend writing specific requests. "Change the date format in column 3 of this file to YYYY-MM-DD" works better than "fix it." Do not hand over a big job all at once. Moving forward and checking one step at a time also saves usage.
MCPY recommends that first-time non-developers start with a copy of the original folder. According to the official quickstart docs, on v2.1.283 and later, auto mode is the default, so Claude Code makes most file edits and runs most commands without asking. Working in a copy you can restore means mistakes cost you little. Change the permission mode with Shift+Tab.
Ways to save usage and cost
With every request, Claude Code sends the whole conversation so far. In a session that has been open for hours, even one short question uses as much usage as the entire conversation. The habits the official cost docs point out are below.
- Clear the conversation with
/clearwhen you move on to a different task. - Sonnet is enough for most work. Use Opus only for hard design problems and switch with
/model. - Turn off MCP servers you do not use (connections that attach external tools) in
/mcp. - For complex work, turn on plan mode with Shift+Tab to confirm the direction first.
- If you take a break and come back, the cache expires and the whole conversation is read again. On a subscription, the cache lasts 1 hour.
What to check when starting in Korea
The pricing page shows prices in dollars, and tax is extra. The pricing page has no official price list in Korean won (KRW). Your charge varies with the exchange rate and VAT. If you assume an exchange rate of 1,400 KRW per dollar, Pro at $20 is 28,000 KRW before tax. This is an arithmetic example and not an actual charge.
A card with overseas payments blocked may be declined for subscription billing. Before you pay, check the overseas payment setting in your card issuer's app. If you are installing on a work laptop, a security policy may block the install command, so ask your IT team first.
Common mistakes
- Trying to log in with a free account. The free plan does not include Claude Code.
- Using it while
ANTHROPIC_API_KEYis set on your computer. According to the official guidance, with this key you are billed API usage rates instead of your subscription. - Reading the dollar figure in
/usageas a charge. It is an estimate that does not apply to subscribers. - Continuing one conversation for days. The longer the conversation, the more usage each question costs.
- Only switching models when you hit a limit. Session and weekly limits stay the same when you change models.
Steps to get started
- Start with monthly Pro billing. Decide on annual billing after you try it for a month.
- After installing, start in a practice folder by asking it to describe the folder.
- Use it for a week and check with
/usagehow close you get to the limit. If you hit it often, consider Max.
Prices and limits can change, so open the pricing page once more before you pay.
Frequently asked questions
Can I use Claude Code for free?
You cannot use it with a free claude.ai account. You need a Pro or higher subscription or a Console (API) account.
Can I use it if I am not a developer?
Yes. Natural-language requests that read and edit files in a folder are enough. If the terminal feels like too much, use the desktop app.
Are the limits separate from Claude chat?
No. On a subscription, Claude and Claude Code share the same usage.
Am I charged automatically when I go over the limit?
No, you have to turn it on yourself. When you hit the limit, you see a prompt to turn on usage credits, and you can open the settings screen with /usage-credits.
How do I update it?
If you installed with the official install script, it updates automatically in the background. With Homebrew and WinGet, you run the upgrade command yourself.
Official source links
- Install and first run: Claude Code quickstart, Advanced setup
- Pricing: Claude pricing page
- Usage and cost: Manage costs effectively, Using Claude Code with Pro and Max plans
- Competing tools: Codex pricing, Gemini CLI quota
Official sources checked: October 1, 2026. The exchange rate example is an arithmetic illustration and not actual billing data. The marketing use examples are estimates adapted from the request style in the official docs. The images are generated illustrations and not actual product screens.
Found this useful? Share it
Related posts
Related projects
Get new posts by email
Insights on marketing, analytics, and dev, delivered to your inbox.

