Claude Code: what it is, what it costs in rand, how to start

Claude Code is Anthropic’s agentic coding tool. You describe what you want in plain language, and it plans the change, edits the files, runs your tests and can open a pull request. It works across your whole codebase, not just the file in front of you.

This guide covers what Claude Code is, who it suits, how to install it, what it costs in rand, and the settings that keep company code safe. Everything was checked against Anthropic’s documentation and pricing pages on 28 September 2026.

What is Claude Code?

According to Anthropic’s documentation, Claude Code reads your codebase, edits files, runs commands and works with your development tools. It runs on several surfaces that share the same engine, settings and project instructions:

The web version also works from the Claude app on iPhone and Android, so you can start a task on your phone and review the pull request later. Anthropic also lists a Slack integration, where you mention @Claude with a bug report and get a pull request back, and GitHub Actions and GitLab CI/CD for automated review.

Who Claude Code is for

Claude Code suits developers who want to hand over whole tasks. Anthropic’s own examples include writing tests for untested code, fixing lint errors across a project, resolving merge conflicts, updating dependencies, writing release notes, tracing a bug from an error message, and reviewing changed files for security issues.

It is harder going if you have never used a terminal or git. The desktop app and the web version lower the barrier, but you still need to read and approve what it changes, because it can change a lot.

How to install Claude Code and sign in

Anthropic lists these system requirements: macOS 13 or later, Windows 10 (version 1809) or later, Ubuntu 20.04 or later, Debian 10 or later, or Alpine Linux 3.19 or later, with at least 4 GB of RAM and an internet connection. It works in Anthropic’s supported countries, which include South Africa.

1. Install it. Anthropic recommends its native installer, which keeps itself up to date in the background. On macOS, Linux or WSL, run:

   curl -fsSL https://claude.ai/install.sh | bash

   On Windows PowerShell, run irm https://claude.ai/install.ps1 | iex. You can also use Homebrew (brew install --cask claude-code) or WinGet (winget install Anthropic.ClaudeCode), but those do not update automatically.

2. Check it worked. Open a new terminal window and run claude --version.

3. Start it in a project. Go to your project folder and run claude.

4. Sign in. On first use, Claude Code sends you to your browser to log in. Use the same account as your Claude Pro, Max, Team or Enterprise plan, or a Claude Console account for pay-as-you-go billing.

One billing trap: if the ANTHROPIC_API_KEY environment variable is set on your computer, Claude Code uses that key instead of your subscription, and you pay API rates. Anthropic’s help centre warns about this, so check before a long session.

Claude Code pricing in rand

Claude Code comes with a Claude plan, or you pay per token through a Console account. Anthropic lists prices in US dollars, excluding tax. Converted at R16.32 to the dollar, before VAT:

Add 15% VAT to the rand figures. On subscription plans, Claude and Claude Code draw on one allowance. When you run out, you can wait for the reset, upgrade, or buy usage credits.

On the API, Anthropic’s prices per million tokens are US$4 in and US$20 out for Opus 5.5 (about R65 and R326), US$2 and US$10 for Sonnet 5 (about R33 and R163), and US$1 and US$5 for Haiku 4.5 (about R16 and R82). Anthropic’s cost guide says enterprise deployments average about US$13 per developer per active day (about R212), or US$150 to US$250 per developer a month (about R2,448 to R4,080), with 90% of users under US$30 (about R490) a day. For the full Claude price list in rand, see Zaiq’s guide to Claude pricing in South Africa.

Practical tips from Anthropic’s documentation

• Clear between tasks. Run /clear when you switch to an unrelated task, because every earlier message is sent again on each turn. Use /compact to summarise a long session instead of wiping it.
• Match the model to the job. Run /model to switch. Anthropic suggests Sonnet for most coding, Opus for hard debugging and wide refactors, and Haiku for quick mechanical edits.
• Point at files instead of pasting them. Name the file path and let Claude read what it needs.
• Keep CLAUDE.md short. This file in your project root holds your coding standards and is read at the start of every session. Anthropic suggests keeping it under about 200 lines.
• Ask for a plan first. Plan mode lets Claude explore the codebase and propose changes before it edits anything.
• Watch your usage. /usage shows your plan usage bars, or an estimated session cost if you use an API key.

Safety settings for company code

Claude Code can run commands on your machine, so set it up on purpose, not by default.

• Know your permission mode. In Manual mode, Claude asks before it edits files or runs most commands. Recent versions start interactive terminal and VS Code sessions in auto mode, where a second model reviews actions and blocks the ones it judges unsafe, instead of asking you. For sensitive repositories, start with claude --permission-mode manual, or set Manual as your default.
• Block secrets. Add deny rules so Claude cannot read files such as .env, for example Read(./.env) under permissions.deny in your project’s .claude/settings.json.
• Keep bypass mode for sandboxes. Anthropic says to use bypassPermissions only in isolated containers or virtual machines where Claude Code cannot cause damage.
• Lock it down centrally. An organisation can deploy managed settings that users cannot override, including permissions.disableBypassPermissionsMode and permissions.disableAutoMode, each set to "disable".
• Use the sandbox. Run /sandbox to limit which files and network hosts Claude’s shell commands can reach.
• Pick the right account. On Free, Pro and Max, you choose whether your coding sessions help train Anthropic’s models; Anthropic keeps data for five years if you allow it and 30 days if you do not. Team, Enterprise and API use is not used for training by default and is kept for 30 days as standard, with zero data retention for qualifying Enterprise customers.
• Mind local copies. Claude Code stores session transcripts on your computer, in plain text, for 30 days by default, so treat developer laptops as holding copies of your code.

If your code, logs or test data contain customers’ personal information, POPIA applies to what you send. Use anonymised test data; our guide on anonymising data before using AI shows how.

Related guides

• Claude AI free: what you get in South Africa, and when to pay
• AI and POPIA: is your data safe with AI tools?
• Chat GPT free? ChatGPT Free vs Go vs Plus in South Africa
• DeepSeek AI in South Africa: is it free, and is it safe?