Pro, Max or the API for Claude Code: choosing by how you work

One tool, three ways to pay for it

Claude Code runs the same whether it's paid for by a subscription or by an API key. What changes is how usage is counted and what happens when you run out. On Pro and Max, Claude Code is included, and all activity in Claude Code and in Claude's chat apps counts against the same usage limits — a long coding session and a long chat draw from one allowance. On Team and Enterprise plans, Claude Code comes with the seat. With an API key, there is no allowance at all: every token bills at API rates, the same rates this site's calculators use. The plans page lists what each plan includes, with Anthropic's list prices.

Check which one you're actually using

The most expensive mistake here is invisible. If the ANTHROPIC_API_KEY environment variable is set on your machine, Claude Code uses that key instead of your subscription. A developer who set the variable months ago for a script, then bought Max, can spend weeks billing API usage while believing they're inside a flat-rate plan. To move back to your subscription, remove the variable from your shell profile, then run /logout, update with claude update, restart your terminal and sign in again with your Claude account.

Choose by how your sessions look

Occasional, bursty work. Pro suits someone who codes with Claude a few times a week and can tolerate the odd pause. As you approach the limit, Claude Code warns you about remaining capacity. When you reach it, Anthropic lists four options: upgrade, turn on usage credits, switch to API billing, or wait for the limit to reset.

All-day work. Anthropic's own guidance is to move to Max 5x if you consistently hit Pro's limits, and to Max 20x if you regularly exhaust Max 5x. A Max plan is billed monthly only, so there's no annual commitment to weigh.

Work that isn't one person at a keyboard. For scripted or scheduled runs, or anything whose cost needs to be attributed to a project or a client, an API key has an advantage no plan offers: every request returns its own usage block, which you can log, total and charge back. The usage tracker turns those blocks into a running spend figure.

Fable changes the calculation

Plans diverge most over the Fable models. Pro leaves them outside its limits, so Fable work there runs on usage credits at API rates; Max folds Fable into the plan's own allowance, capped at a share of the weekly limit. So if you expect to lean on Fable for hard problems, Max treats that as part of the plan and Pro treats it as metered usage — see using Fable on a Pro plan for pricing the metered version.

Before assuming you need Fable at all: Anthropic's guidance treats Opus 5.5 as the right model for most work — complex agentic coding included — and keeps Fable 5.1 for the hardest long-running jobs. Most Claude Code sessions fall on the Opus side of that line.

Making a plan go further

Effort. Claude Code runs Opus 5.5 and Sonnet 5.5 at medium effort by default. Dropping to low for routine edits stretches an allowance; raising effort is for the problems that need it. See Claude Code effort levels for how to change it per session.

Plan with Opus, build with Sonnet. The opusplan model alias uses Opus while you're in plan mode and switches to Sonnet to carry the plan out — Opus's reasoning for the plan, Sonnet's efficiency for carrying it out. Plan mode covers how it works.

Fresh context for a fresh task. Every turn works over the whole conversation so far. When you switch to an unrelated task, /clear starts it without carrying the old history along.

Usage credits: the overflow valve

A plan with usage credits turned on behaves like a flat rate up to your limit and API billing beyond it. Credits bill at standard API rates, and one balance covers Claude Code, Claude's apps, Cowork and third-party products. That makes Pro plus credits a genuine middle option for someone whose usage is mostly light with occasional heavy weeks — cheaper than Max in a light month, and still able to keep working when a deadline pushes past the limit.

Deciding with your own numbers

Put a typical month through the subscription vs API cost calculator: your usage shape gives the API cost, and each plan's list price is one tap away to compare against. If the API figure sits well under a plan's price, an API key or Pro plus credits is likely cheaper. If it sits well over it, the subscription is doing its job.