Claude Code · pricing guide · vibe coding

ShipAI.today engineering guide

Claude Code pricing in 2026: plans, token costs, and what you'll actually pay.

Claude Code pricing is not straightforward — it's tied to your Claude.ai subscription tier or Anthropic API usage. This guide breaks down every plan, explains token costs, and tells you which option makes sense for a SaaS founder doing serious vibe coding.

4 pricing options comparedToken cost calculator breakdownSetup guide included

See all plans Setup guide

$20

Pro (limited)

$100

Max 5× plan

$200

Max 20× plan

~$3–15

Per MTok (API)

Plans

How Claude Code pricing works — every option

Claude Code is not sold as a standalone product. Access is bundled with Claude.ai subscriptions or billed directly via the Anthropic API.

The key thing to understand about Claude Code pricing

Claude Code is a CLI agent — not a separate app you subscribe to. You access it either via your Claude.ai subscription (which sets a usage quota) or by providing your own Anthropic API key (pay-per-token). Both approaches give you the same Claude Code CLI — the difference is how usage is metered and billed.

Plan	Price	Claude Code access	Models	Best for
Claude.ai Pro	$20/mo	Limited — included but rate-capped	Claude Sonnet (primary)	Occasional Claude Code use — light tasks, evaluation
Claude.ai Max (5x)	$100/mo	Full — 5× the Pro quota	Claude Sonnet + Opus access	Solo SaaS founders doing daily autonomous coding runs
Claude.ai Max (20x)	$200/mo	Full — 20× the Pro quota	Claude Sonnet + Opus — heavy usage	Power users running multi-hour autonomous refactors daily
Anthropic API (BYOK)	Pay-per-token	Full — no quota cap, direct API billing	Claude Opus 4, Sonnet 4, Haiku	Developers who want usage-based billing and full model control

API token costs

Claude Code token pricing — what you pay per request

If you use the Anthropic API directly (BYOK mode), you pay per million tokens. Here's what each model actually costs and how to estimate your bill.

Model	Input cost	Output cost	Typical monthly (BYOK)	When to use
Claude Opus 4	$15 / MTok	$75 / MTok	$30–120/mo heavy use	Best reasoning — use for complex refactors only
Claude Sonnet 4	$3 / MTok	$15 / MTok	$8–35/mo daily use	Best daily driver — fast, capable, affordable
Claude Haiku 3.5	$0.80 / MTok	$4 / MTok	$2–8/mo	Good for fast, simple completions and summaries

Estimating your Claude Code bill (API mode)

A typical Claude Code session (1 hr, medium complexity) ≈ 200k–500k tokens = $0.60–$7.50 with Sonnet 4
A long autonomous refactor (4+ hrs) ≈ 1M–3M tokens = $3–$45 with Sonnet 4
Claude Code compresses context intelligently — actual costs are often lower than naive estimates
Monitor spend in real time: anthropic.com/account → Usage dashboard

Max plan vs API billing — which is cheaper?

Max $100/mo includes 5× Pro quota — roughly equivalent to $3–15 of API spend/day
If your typical day uses less than $3 of API tokens: API billing is cheaper
If you run multi-hour autonomous sessions daily: Max $100 or $200 is more predictable
API billing is better for teams — each developer's usage billed separately, no shared pool

Setup guide

How to set up Claude Code for SaaS development

Get Claude Code running in your project in under 10 minutes. Written for Next.js / TypeScript SaaS developers.

Install Node.js 18+ and npm

Claude Code requires Node.js 18 or higher. Verify with: node --version. If you're on macOS, use brew install node or nvm.

node --version  # needs 18+
npm install -g @anthropic-ai/claude-code

Authenticate with Anthropic

Run claude from your project directory. On first run, it opens a browser window to authenticate with your Anthropic account (Claude.ai subscription or API key).

cd your-project
claude  # opens auth flow on first run

Set up CLAUDE.md for your codebase

Claude Code reads CLAUDE.md at the repo root as persistent context. Add your tech stack, conventions, and patterns here. This dramatically improves output quality.

# CLAUDE.md
Stack: Next.js 15 App Router, TypeScript, Drizzle, Better Auth, Stripe
Patterns: All DB queries use Drizzle. Auth handled via Better Auth sessions.
Conventions: Use server components by default. Client components must be marked 'use client'.

Run your first task

Claude Code is a CLI agent — give it a task and it runs autonomously. Use --dangerously-skip-permissions to avoid approval prompts on trusted codebases.

claude 'Add a usage tracking table to the Drizzle schema and update the billing middleware to enforce limits'

Use CLAUDE.md to get 10× better results

Claude Code reads CLAUDE.md at your repo root as persistent context before every session. This is the single biggest quality multiplier. Include: your tech stack, naming conventions, folder structure, which patterns to follow (and which to avoid), and any gotchas in your codebase. Think of it as a permanent system prompt that makes Claude Code behave like a senior developer who already knows your project intimately.

How to use Claude Code

Claude Code patterns that work well for SaaS development

Practical examples from vibe coding a production SaaS — what to delegate, how to prompt, and when to step in.

Build a feature end-to-end

Excellent

Let Claude Code plan and execute — give it the full spec upfront.

claude 'Build a team invite system: DB table, invite email via Resend, invite acceptance flow, and update the dashboard to show pending invites.'

Fix failing tests

Excellent

Claude Code runs the suite, reads failures, patches, re-runs. Fully autonomous.

claude 'Run the test suite, identify all failures, and fix them. Do not change test intent — fix the implementation.'

Migrate a library or schema

Excellent

Schema migrations with Drizzle generate, run, query updates — all in one run.

claude 'Add a usage_events table to the Drizzle schema, generate and run the migration, update the billing middleware to log events, and add a query helper.'

Exploratory debugging

Good

Less autonomous but useful — Claude Code reads logs and traces errors across files.

claude 'The Stripe webhook handler is failing with a 500. Read the logs, trace the issue through the codebase, and fix it.'

UI component building

Good

Works but Cursor with @docs and inline preview is faster for UI iteration.

claude 'Create a responsive pricing table component matching the existing design system. Use the NeoPanel component and Tailwind conventions from CLAUDE.md.'

Documentation writing

Excellent

Claude Code can read your codebase and generate accurate CLAUDE.md or README updates.

claude 'Read the src/lib/billing directory and write a detailed explanation of how billing and usage limits work for CLAUDE.md.'

FAQ

Claude Code pricing questions answered

How much does Claude Code cost per month?

It depends on how you access it. Via Claude.ai Pro ($20/mo) you get limited Claude Code access. The Max plan at $100/mo or $200/mo gives 5× and 20× more quota respectively. If you use your own Anthropic API key (BYOK), you pay per token — typically $8–45/mo with Sonnet 4 for daily use. Heavy autonomous workflows can cost $100+/mo in API fees.

Is there a free tier for Claude Code?

No. Claude Code requires either a Claude.ai paid subscription (starting at $20/mo) or an Anthropic API key with billing set up. There is no free quota for Claude Code — unlike Cursor's 50 free requests or GitHub Copilot's free tier.

What is Claude Code Max and is $100/mo worth it?

Claude.ai Max at $100/mo gives you 5× the usage quota of Pro. For a solo developer running Claude Code for 2-4 hours/day, Max 5× is usually sufficient. Max 20× ($200/mo) is for power users running multi-hour autonomous builds daily. If you're hitting quota limits on Pro regularly, upgrade to Max instead of throttling your workflow.

How do I set up Claude Code?

Install via npm: npm install -g @anthropic-ai/claude-code. Then run claude from your project directory — it will prompt you to authenticate via your Anthropic account or API key. Create a CLAUDE.md file at the repo root with your stack and conventions. That's it — you're ready to run tasks.

Should I use Claude Code with a subscription or my own API key?

If you use Claude Code occasionally (less than 1 hour/day), a Claude.ai subscription is simpler and predictable. If you run heavy autonomous sessions or prefer per-usage billing with full model access (Opus 4 vs Sonnet), the API key approach gives more control. Teams especially benefit from BYOK since each developer's costs are tracked independently.

How does Claude Code compare to Cursor in cost?

Cursor Pro is $20/mo flat — unlimited completions, 500 fast requests. Claude Code is effectively $100+/mo for serious use (Max plan) or variable API costs. Cursor is the better value for daily in-IDE work. Claude Code wins for long autonomous tasks where you're OK paying more per run. Many developers use both: Cursor for daily coding, Claude Code for big migrations or feature builds.

ShipAI.today's codebase is designed for Claude Code to navigate.

A well-organized repo with a detailed CLAUDE.md, clear module boundaries, and consistent patterns means Claude Code spends less time understanding structure and more time writing useful code. Pre-wired auth, billing, and AI modules reduce the scope of every task.

Ships with a pre-written CLAUDE.md — Claude Code understands the stack from session 0
Drizzle schema + Better Auth + Stripe billing all pre-wired — less scaffolding per task
Consistent Next.js 15 App Router patterns that Claude Code can follow without correction

Get ShipAI.today More vibe coding guides