What a Claude plan buys compared with API list prices

On 18 October 2026, Anthropic's pricing page lists Claude Pro at $20 a month billed monthly, or $17 a month on annual billing. The Claude API, by contrast, charges per million tokens. The two are priced in different units, so comparing them takes a little arithmetic and a clear view of what each side does not tell you.

This post uses only Anthropic's own pages, fetched on 18 October 2026. Prices change, so check the links at the bottom before you decide anything.

What the plans cost

From claude.com/pricing:

What the plans do not publish

Every plan has limits that reset on a rolling five hour session window, and paid plans add weekly limits on top. Anthropic's support article on Pro and Max usage does not publish token counts or message quotas. It describes limits only as time windows, shared between Claude and Claude Code.

That means nobody outside Anthropic can state how many tokens a Pro plan includes. Any article that gives you a token figure for a subscription is estimating. This one will not.

When you hit a limit, Pro and Max users can turn on usage credits to keep working at standard API rates, according to Anthropic's pricing FAQ. At that point you are paying list price for the extra use.

The API side

The platform pricing page lists prices per million tokens (MTok). A few current rows:

The Batch API takes 50 percent off input and output. Claude 4.7 and later models use a newer tokenizer that Anthropic says produces about 30 percent more tokens for the same text, so token counts are not comparable across model generations.

Turning a plan price into list price terms

You can ask a fair question without knowing the plan's limits: how much work does $20 buy at list price? Using Sonnet 5.5 at $2 input and $10 output per MTok:

• $20 buys 10 million uncached input tokens, or 2 million output tokens.
• With cache reads at $0.10, $20 buys 200 million cached input tokens.

Those are list price facts. They do not say what a Pro plan allows, because Anthropic does not publish that. They tell you the other side of the comparison: what the same money buys on the API.

Try it yourself

Pick a typical working day and count the sessions. Then run this to see the API list price of one session. Change the numbers to match yours.

# Sonnet 5.5 list prices, USD per million tokens, from platform.claude.com on 18 Oct 2026
IN, OUT, CACHE_READ = 2.00, 10.00, 0.10

def cost(input_tokens, output_tokens, cached_input_tokens=0):
    return (input_tokens * IN + cached_input_tokens * CACHE_READ
            + output_tokens * OUT) / 1_000_000

one_session = cost(input_tokens=200_000, output_tokens=20_000, cached_input_tokens=800_000)
print(round(one_session, 2), "per session")
print(round(20 / one_session, 1), "sessions equal one month of Pro at $20")

Compare that figure with how many sessions you really run in a month. If you run far fewer, the plan price is likely the larger cost. If you run far more, check whether you hit the plan limits, because those limits are what Anthropic's page says governs what you get. Our guide to estimating the tokens in one session shows how to get the inputs right.

What this means for a team

Plan, API and Enterprise are three different billing shapes: a flat fee with limits, pay per token with no ceiling, and a seat fee plus API rate usage. Safeguard supports bringing your own model, and its MCP tool calls are unlimited, so the platform side of your security tooling does not add a per call meter to whichever model bill you choose.

Quick check

Sources

• Claude pricing, fetched 18 October 2026
• Claude API pricing, fetched 18 October 2026
• Using Claude Code with your Pro or Max plan, fetched 18 October 2026