"Setting up Claude Code for side-income coding" has been a popular topic in online coding communities lately. The specific results people report vary a lot and are hard to verify from the outside. This article addresses the real question behind that interest — how do you actually install Claude Code, which plan should you pick, and how do you use it for real coding work — using only what's confirmed in the official docs (code.claude.com, support.claude.com). Following the steps below, from installation through cost management, takes roughly 30-40 minutes to get a working setup.
🟢 Model references match the current lineup · model notice · Fable subscription
Fable 5 and 5.1 subscription (updated September 7, 2026): Claude Fable 5.1, released September 1, 2026, is the current Fable model and Fable 5 is now legacy. Plan terms are the same for both — Max and Team Premium plans include Fable at up to 50% of the weekly usage limit; Pro and Team Standard use usage credits (
What you'll be able to do after this guide
By the end, you'll have Claude Code (Anthropic's official CLI — command-line interface — tool for AI-assisted coding in the terminal) installed, logged in, and running its first session; a CLAUDE.md file set up to carry project context; and a working understanding of plan costs before you start real work. Installation through your first code change usually takes about 30-40 minutes.
One caution up front: "setup" only prepares the tool. How much you actually earn from freelance work, side projects, or selling automation scripts with it depends heavily on your skill, time invested, and market demand. This article makes no income promises — it focuses only on setting the tool up correctly.
Before you start
- A computer where you can open a terminal (command-line window) — macOS, Linux, and Windows (including WSL — Windows Subsystem for Linux) are all supported. If you've never used a terminal, check the official terminal guide first.
- A paid Claude subscription (Pro, Max, Team, or Enterprise) or a Claude Console (pay-as-you-go API) account — Claude Code is not available on the Free plan. The official quickstart guide lists Pro, Max, Team, Enterprise, or Console as the required account types; Free isn't among them. Starting with Pro ($20/month) is the usual entry point.
- A code project folder to work in — new or existing, either works.
Quick terms
- CLI (Command Line Interface) — a program you run with typed commands in a terminal. Claude Code isn't an app you click through; it's a CLI tool you drive with terminal commands.
- Token — the smallest unit of text an AI model processes. Pay-as-you-go API costs are billed by token usage. On a subscription (Pro/Max), you pay a flat monthly fee, so you don't need to track token pricing directly.
- CLAUDE.md — a project instructions file. It's read automatically at the start of every session so you don't have to re-explain the same context each time.
- MCP (Model Context Protocol) — a standard for connecting external tools and services. Used to connect things like GitHub or a database to Claude Code.
- Subagent — a helper Claude instance dedicated to one task. Useful for offloading verbose work like running tests or scanning logs so the output doesn't clutter your main conversation.
- Plan mode — a mode where Claude proposes an approach before editing anything. Useful before complex tasks, to catch a wrong direction before it costs you rework.
Step 1: Install
Run the command for your OS in a terminal. The official docs list these methods:
| Environment | Install command | Notes |
|---|---|---|
| macOS / Linux / WSL | | Officially recommended. Auto-updates in the background. |
| Windows PowerShell | | Your prompt starts with PS C:\ if you're in PowerShell. |
| Windows CMD | | Your prompt shows C:\ without PS if you're in CMD. |
| macOS (Homebrew) | | Doesn't auto-update — run brew upgrade claude-code yourself. |
| Windows (WinGet) | | Also doesn't auto-update. |
Debian, Fedora, RHEL, and Alpine also support installing via their own package managers (apt, dnf, apk) per the official docs. Once installed, run claude --version. You're set if it prints a version number followed by "(Claude Code)".
Step 2: Log in and open your first session
Type just claude in your terminal, and the login flow starts automatically.
claudeFollow the browser prompts to complete login. To switch accounts later, type /login inside a running session. Once logged in, you'll see a prompt showing the current version, model, and working directory — that screen means login succeeded.
On Pro, Max, and Team plans, new sessions have started in "auto mode" since August 14, 2026. In this mode a classifier (a separate model that judges how risky an action is) reviews actions in your place, flagging only the risky-looking ones and letting other file edits and commands run without asking. If you set your own default in ~/.claude/settings.json, that setting is kept, and an organization-managed default is not changed either. As of this writing, auto mode is opt-in on Enterprise plans and via the Claude API, so those start in "Manual mode," which asks every time. Press Shift+Tab to cycle until the mode you want appears in the status bar.
Caution: auto mode is convenient, but on real client work or production code, make it a habit to review changes with git status and git diff before committing. Skipping that review lets unwanted changes slip through.
Step 3: Set project context with CLAUDE.md
CLAUDE.md is an instructions file you place in your project folder; it's read automatically at the start of every session. Fill it with your project's purpose, coding conventions, and frequently used commands so you don't repeat the same explanations every time. Asking Claude in plain language to "analyze this project and create a CLAUDE.md file" gets you a draft based on the actual project structure. The official docs recommend keeping CLAUDE.md under 200 lines with only the essentials — a bloated file adds unnecessary content to every session's context, which costs more.
Step 4: Learn the real coding workflow
Once the basics are set up, you just describe tasks in plain language. Here are the commands you'll use most:
| Command | What it does |
|---|---|
claude | Start an interactive session |
claude "task" | Start a session with an initial request |
claude -p "query" | Ask once, then exit |
claude -c | Continue the most recent conversation in this folder |
claude -r | Pick a past conversation to resume |
/clear (in-session) | Reset conversation history — important for cost control |
/help (in-session) | List available commands |
A typical workflow: first let Claude explore the codebase ("what does this project do?"), then make a specific request ("add input validation to the login form"), review the resulting changes, and wrap up with something like "run the tests" or "commit these changes." The official docs recommend being specific — "fix the bug where users see a blank screen after login" works better than "fix the bug." Specific requests mean Claude searches fewer irrelevant files, which is both faster and cheaper.
Managing costs: choosing a plan and saving habits
If you're using this for side income, cost control matters. The short answer: start on a subscription plan (Pro or Max) and upgrade only once you're consistently hitting limits — it's the most predictable option for an individual.
- Pro ($20/month) — usage resets every 5 hours within a session limit. A reasonable starting point.
- Max 5x ($100/month) and Max 20x ($200/month) — 5x and 20x more usage per session than Pro, respectively. On top of the session limit, a separate weekly limit resets at a fixed time each week.
- Claude Console (pay-as-you-go API) — billed by tokens used. Official docs cite an average of about $13 per developer per active day and $150-250 per month across enterprise deployments, with 90% of users under $30 per active day. That figure applies to metered API billing, though — on a subscription (Pro/Max) you're paying a flat fee, so it doesn't translate directly.
As of this writing, a promotion increasing the weekly usage limit by 50% for Pro, Max, and Team plans is running through August 31, 2026 (Pacific time). It has been extended multiple times before, so check the official usage limits guide to confirm whether it's still active before relying on it.
The official docs recommend five saving habits. ① Pick the right model — Sonnet handles most coding tasks and costs less than Opus, so switch to Opus (/model) only for complex architectural decisions. ② Clear context proactively — run /clear when moving to unrelated work so you're not paying to reprocess stale history. ③ Trim MCP overhead — check connected external tools with /mcp and disable servers you're not using. ④ Delegate to subagents — hand off verbose work like running tests or scanning logs so the noisy output stays out of your main conversation. ⑤ Use plan mode — press Shift+Tab before complex tasks so Claude proposes a direction first, cutting the rework cost of going down the wrong path.
Common snags and fixes
- "The token '&&' is not a valid statement separator" on Windows — happens when you run a CMD command in PowerShell. If your prompt starts with
PS C:\, use the PowerShell command (irm ... | iex) instead. - "'irm' is not recognized as an internal or external command" — the reverse: a PowerShell command run in CMD. If your prompt is
C:\withoutPS, you're in CMD, so use the CMD command. claude --versiondoesn't work after install — open a fresh terminal window so your PATH picks up the new install.- A "You've hit your weekly limit" message — check the reset time shown on screen. If you need to keep working before then, run
/usage-creditsinside a session to request additional usage. Some versions can also wait automatically until the limit resets and resume your task. - Usage draining faster than expected — common causes include leaving a session open for a long time, a long break causing a cache miss (so the full context gets reprocessed), or sticking with Opus as the default model. Run
/usagein a session to see recent usage and hints about the cause.
Going further
Once the basics feel comfortable, a few next steps are worth trying. For a task you repeat often (like PR reviews or database migration checks), move the instructions out of CLAUDE.md into a skill so it only loads on demand, keeping your base context lighter. For large tasks touching many files, leaning on subagents and plan mode cuts down on rework. For the full picture on plans and limits, see the usage limits guide.
Frequently asked questions
Q. Can I use Claude Code for free?
No, there's no free tier for it. The official quickstart lists Pro, Max, Team, Enterprise, or a Claude Console (pay-as-you-go) account as the required options — Free isn't among them. Most individuals starting out begin with Pro ($20/month).
Q. Can I still install it via npm?
The current official quickstart only documents the curl script (macOS/Linux/WSL), PowerShell/WinGet (Windows), and Homebrew (macOS) — it doesn't mention npm. If you're installing fresh, follow the methods listed there.
Q. If I hit a usage limit, am I locked out entirely?
No, limits reset automatically at a scheduled time. If you need to keep working sooner, run /usage-credits inside a session to request additional usage, or wait until the reset time shown on screen.
Q. Does this setup guarantee I'll make money coding?
No, setup alone doesn't guarantee income. Claude Code is a tool that helps with coding tasks; actual results depend on what you build, how much time you invest, and whether there's market demand. This guide only covers setting the tool up correctly.
Related articles
- How to Prompt Claude Well: 7 Basics of Writing Prompts
- Claude Mobile App Deep Dive — Widgets, Voice Mode, and Device App Integration
- Claude Code Installation Guide: native vs npm, Windows Included
- Claude Code Skills (Agent Skills) Guide: Extend Claude with SKILL.md
- Claude Code MCP Setup: claude mcp add, Scopes, and Transports