TL;DR
- Claude Code runs on Claude subscriptions: Pro ($20/mo), Max $100 (5x usage), Max $200 (20x usage).
- For most freelancers and small teams: Claude Pro $20/mo covers daily Claude Code work.
- TCO per year: Claude Pro $240, Cursor Pro $240, GitHub Copilot Pro $120. Cursor and Claude Code same price, different paradigms.
- API pay-per-token makes sense for companies building custom systems with Claude, not for developers using Claude Code as IDE assistant.
- Recommendation: start with Pro, upgrade to Max $100 when you hit daily limits during active work.
Claude Code pricing isn't a simple table: Claude Code doesn't exist as a standalone product with its own price sheet, but as a feature built into Claude subscriptions and Anthropic API access. This post breaks it down.
Claude Code pricing 2026: quick answer
Claude Code is available in all paid Claude plans, starting at $20/month.
Anthropic.com/pricing offers three subscription tiers and API access:
- Claude Pro ($20/mo): Claude Code with standard usage limits. Sufficient for daily work if you don't use Claude Code 8+ hours per day.
- Claude Max $100/mo: 5x more usage limits than Pro. For developers who intensively use Claude Code for autonomous tasks throughout the day.
- Claude Max $200/mo: 20x more than Pro. For power users and companies where Claude Code is a critical production tool.
- API (pay-per-token): For companies building custom systems, integrations, scripts. No monthly subscription, you pay for consumed tokens.
Check current prices directly on Anthropic's site before purchasing: Anthropic regularly updates plans.
How does Claude Pro differ from Claude Max for Claude Code?
The only difference between Pro and Max is usage limit: how many tokens you can process per day.
Claude Code in Pro and Max uses the same models (default Claude Sonnet 4.6, optionally Claude Opus 4.8 for harder tasks). There's no restriction on task types: refactoring, writing tests, autonomous agents, working with large files, all identical in both plans.
The difference shows up with intensive use. When you give the agent a complex task (e.g., "read all files in this directory, write unit tests for each class, run them and fix errors"), one session can consume hundreds of thousands of tokens. On Pro you can hit the daily limit with a few such sessions per day.
Heuristic: if you use Claude Code less than 4 hours daily and don't assign long autonomous tasks, Pro suffices. If Claude Code is your primary tool throughout the workday, Max $100 eliminates frustration with limits.
When does Claude Max $100 make sense?
Claude Max $100 pays off when Pro's limit interferes with your work more than once per week.
Concrete situations where Max $100 is justified:
- You're writing a new application from scratch and Claude Code works autonomously through multi-hour sessions.
- You're doing large refactors where the agent reads hundreds of files to understand context before each change.
- You use Claude Code to generate documentation from code (input + output tokens can be massive with large codebases).
- You work for multiple clients simultaneously and switch between projects all day.
For agencies or companies: Max $100 per developer actively using Claude Code is a reasonable business line. At a billable rate of $35/hour (≈150 PLN/h) and 3h/week savings, ROI is 12h × $35 = $420 at a cost of approximately $100 monthly.
Is Anthropic API cheaper than subscription?
API pay-per-token is cheaper per token than subscription, but isn't designed for interactive Claude Code work in the terminal.
Claude Code as a CLI tool (launched with claude command in terminal) runs through Claude subscription. API is for developers building custom applications using Claude models: chatbots, automations, document processing systems, integrations with proprietary code.
When API makes sense instead of subscription:
- You're building a script that automatically processes hundreds of files with Claude and run it once weekly, not interactively.
- You're integrating Claude with your own platform or CRM.
- You're building a system with multiple parallel sessions for different clients.
With API you pay per token: input and output separately. At high volumes (millions of tokens monthly) costs can be lower than Max $200, but require managing your own usage and lack the comfort of fixed-cost subscription. Model pricing details at Anthropic pricing.
Claude Code vs Cursor vs GitHub Copilot: TCO per year
Total cost of ownership over 12 months for one developer:
| Tool | Plan | Monthly | Year | Payment |
|---|---|---|---|---|
| Claude Code | Pro | $20 | $240 | Claude subscription |
| Claude Code | Max $100 | $100 | $1,200 | Claude subscription |
| Cursor | Pro | $20 | $240 | separate subscription |
| GitHub Copilot | Pro | $10 | $120 | GitHub billing |
| GitHub Copilot | Pro+ | $19 | $228 | GitHub billing |
Cursor Pro and Claude Pro are the same price ($20/mo, $240/year), but different paradigms: Cursor is IDE-first with project context in VS Code, Claude Code is terminal-first agent working autonomously. You can use both simultaneously because they solve different problems, more in tool comparison.
GitHub Copilot Pro is cheapest ($120/year), but has the weakest project context of the three. Copilot Pro+ ($228/year) is closer to Cursor in price and model quality.
Decision: if you work solo and want one tool, Cursor Pro or Claude Pro depending on work style. If you want both approaches (IDE + terminal agent), $40/mo for both Pro plans is still less than one day's work.
Claude Code pricing: which plan to choose?
Depends on usage intensity and work style:
Claude Pro ($20/mo): For freelancers and developers who use Claude Code a few hours daily on medium-scale projects. This is the entry point and sufficient for most through the first months.
Claude Max $100/mo: For developers who have Claude Code as their primary production tool all day and regularly hit Pro's limit. Upgrade when limits interfere, not "just in case."
Claude Max $200/mo: For power users or small teams where one account handles intensive use. Rarely needed for individual developers.
API: For companies building custom products with Claude, not for personal use of Claude Code as assistant.
Practical approach: start with Pro. If you don't hit the limit over 2 weeks, Pro suffices. If you hit the limit during normal work, upgrade to Max $100 and test for a month. Don't overpay "for reserve."
Claude Code free: can you use it without payment?
Claude Code is not available in Claude's free plan.
The free Claude.ai plan gives access to the browser chatbot with daily limits, but doesn't include access to Claude Code as a CLI tool. To run claude in terminal and use Claude Code autonomously, you need minimum Claude Pro.
Exception: if you have API access through an API key (e.g., through company integration or your own project), you can run Claude Code with your own API key and pay per token instead of through subscription. This is an option for developers who already have an API account with billing enabled.
To get started, check Claude Code documentation and choose a plan at Anthropic.com/pricing.
If you're considering implementing AI agents in your company or want to build a custom workflow based on Claude Code, check our AI services or browse the agents hub.