Is Claude Code included with Claude Pro?

Is Claude Code included with Claude Pro?

Yes. Claude Code is included with Claude Pro, with Max, with Ultra, with every Team seat and with current Enterprise seats, so one subscription covers the chat apps and the terminal tool. On an older Enterprise contract it depends on the seat type: only Premium seats and Chat + Claude Code seats include it, so check which seat you hold. The free Claude plan does not include it. Anthropic's setup documentation is explicit: Claude Code requires a Pro, Max, Ultra, Team, Enterprise or Console account, and the free plan does not include Claude Code access. A subscription needs no API key (application programming interface key, billed per token). A key is optional: if ANTHROPIC_API_KEY is set and you approve it, that session is billed per token instead, which suits automated runs you want billed separately from your interactive work, or an occasional session you choose to bill to a specific key. Without a subscription, a Console account is the route, and it bills per token.

The part people get wrong comes next. Claude Code has no separate allowance of its own. It spends the same subscription usage budget your chat conversations spend, so a long coding session shortens the chat window that follows it. Anthropic states that Pro, Max and Ultra plans offer usage limits that are shared across Claude and Claude Code, meaning all activity in both tools counts against the same usage limits. The one exception is a current usage-based Enterprise plan, which has no included budget to share: there are no per-seat usage limits, and every request in either tool is billed at API rates.

What each plan actually gets

• Free: no Claude Code. Chat on the web and in the apps, nothing in your terminal.
• Pro: Claude Code included at Pro usage levels. $20 per month in the United States as of October 2026, lower on annual billing.
• Max: the same Claude Code with five times Pro usage every five hours and every week, at $100 per month on the web as of October 2026, billed monthly only. Includes $100 a month in API credits.
• Ultra: the same Claude Code with ten times Pro usage every week, at $200 per month on the web as of October 2026, billed monthly only. Includes $200 a month in API credits. Mobile prices for Max and Ultra may vary.
• Team and Enterprise: included with every Team seat, where the allowance belongs to the seat and a premium seat carries more of it than a standard seat. Current Enterprise plans use a single seat type that includes Claude Code, with no per-seat allowance: all usage is billed at API rates, under spend limits an admin sets. Older Enterprise contracts use separate seat types: only Premium seats and Chat + Claude Code seats include Claude Code, and only seat-based contracts have an included allowance.

Those are the only prices here, because plan prices move and this page does not. The current numbers live in a full cost breakdown for Claude Code, the choice between plans is worked through in a plan by plan comparison of the Claude subscriptions, and what the free Claude tier does cover is worth checking before you assume you are stuck.

Why one coding session shortens your chat window

Two windows govern a Pro, Max or Team subscription, and a seat on an older seat-based Enterprise contract. A session limit resets every five hours. A weekly limit resets at a fixed time each week that is assigned to your account, and you can see your own reset time in the usage screen. Verified Ultra accounts have no five-hour limit; unverified accounts have a temporary limit. Ultra still has a weekly limit across all models, and Max and Ultra cap Fable at half of the regular weekly allowance, drawing from that same budget and using it faster than other models. A current usage-based Enterprise plan has neither window: it bills consumption, and admins cap it with spend limits instead. Neither window is a message count you can calculate in advance, because how far you get depends on how long your conversations are and which model you run. Treat them as behaviour to watch, not as a number to budget against. The limit mechanics in detail are worth reading once, before you hit one.

Claude Code drains that budget faster than chat because of what a coding turn carries. Every request sends the whole conversation so far. Each tool call, a file read or a test run, sends another request carrying the results of the previous one. Anthropic's own guidance for administrators says to budget more for a coding seat than a chat seat for exactly this reason: one debugging session can consume more than a day of chat.

Prompt caching softens this and does not remove it. Repeated context is re-read at a cached rate, so a one-line question inside a session that has been open all day still draws usage for the entire conversation behind it. On a subscription, while you are within your plan's included usage, the main conversation gets a one-hour cache lifetime; subagents, compaction and most other requests Claude Code makes outside that conversation default to five minutes. Step away for longer than an hour and your first message back misses the cache, which means your full context is written to the cache again at the cache-write rate. That rate is above the normal input rate, not equal to it: Anthropic prices one-hour cache writes at 2 times the base input price and five-minute cache writes at 1.25 times. That one message can cost more than the twenty before it.

The practical signs you are burning the shared budget

Run /usage inside Claude Code. On a paid plan it draws usage bars for your plan limits, with a breakdown of what has been spending them. Press d for the last 24 hours and w for the last 7 days.

The breakdown flags any behaviour responsible for 10 percent or more of recent usage, such as long context or cache misses, and it attributes usage to skills, subagents, plugins and individual MCP (model context protocol) servers. One caveat matters. Those figures are computed from local session history on the machine you are sitting at, so work you did on another machine or on claude.ai is not in the breakdown. The plan bars are global. The attribution is local.

When you run out, the message names which ceiling you hit:

You've hit your session limit
You've hit your weekly limit

Both of those windows apply across every model, so switching model with /model does not give you access back. A model-specific message such as You've hit your Opus limit is a different thing, and there /model does keep you working, on another model. A context warning or an auto-compact notice is not a usage limit at all: it means the conversation is near the session's context window and older history is about to be summarised. Anthropic's cost guidance puts context management first, and keeping the context window small has the clearest effect on how fast the budget drains. /clear between unrelated tasks costs nothing and starts the next request with less to carry. If you are already locked out, the options when a limit is reached come down to waiting, upgrading, or turning on usage credits. If Anthropic has given your account a free limit reset, that is a fourth option: use it from Settings > Usage on the web or in Claude Desktop (not from Claude Code itself), and the session or weekly limit it covers is back to full in Claude Code as well.

Why an API key can take over the bill

This one bites hardest on a server. If the ANTHROPIC_API_KEY environment variable is set and you approve it, Claude Code authenticates with that key instead of your subscription, and the work is billed per token to your API account. In an interactive session it asks once whether to use the key and remembers the answer, so a key you approved months ago keeps taking over without another prompt; to change that choice later, use the "Use custom API key" toggle in /config, which appears only while the variable is set. In non-interactive mode (claude -p) it uses any key that is present, with no prompt at all. Your Pro, Max or Ultra allowance sits untouched while the invoice grows. There is no free allowance waiting on the other side either, because the Claude API has no free tier beyond the small credit granted at signup, and the monthly API credits that come with Max, Ultra and Team plans do not cover interactive Claude Code. They do cover claude -p runs and Agent SDK work you start yourself with an API key from the Console organization linked to your plan.

On a VPS this happens by accident more often than on a laptop, because the same shell profile usually holds keys for other tools. Anthropic's own advice is to keep ANTHROPIC_API_KEY unset if you want your subscription to pay, and to set a key only for the terminal session that needs it. After you log in, Claude Code also shows a notice when a subscription login and an API key are both active. Check before you start:

printenv ANTHROPIC_API_KEY

Nothing printed means no API key is set in this shell. That alone does not prove Claude Code will use your subscription: a credential can also come from ANTHROPIC_AUTH_TOKEN, an apiKeyHelper script, or the env block of a settings file, including settings an administrator manages, and any of those takes precedence over a claude.ai login. Run /status inside Claude Code to see which one is active. A key printed means the next session bills the API if you approved that key before or run claude -p; otherwise Claude Code asks first, and declining keeps you on the subscription. Clear it for this shell, then find the file that keeps setting it:

unset ANTHROPIC_API_KEY
grep -rn ANTHROPIC_API_KEY ~/.bashrc ~/.profile ~/.bash_profile 2>/dev/null

Now start Claude Code and sign in with the subscription account:

claude

The first run starts a browser login. Inside a session, /login switches accounts, /status names the authentication method in use, and /usage confirms which billing you landed on: a subscriber sees plan usage bars, while an API account sees a session block of token costs and no plan bars. The rest of the server-side setup, including how to keep a session alive after you disconnect, is covered in running Claude Code on a VPS inside tmux, and the permission side of it in running Claude Code safely on a server you own.

One related trap. When the plan allowance runs out, you can turn on usage credits to keep working past it, billed on top of the subscription. Once you are drawing on credits, Claude Code drops the main conversation's prompt cache lifetime from one hour to five minutes, so the same idle gaps that used to be free start costing you, unless you set the promptCacheTtl setting to 1h yourself. /usage-credits opens Settings > Usage for a Pro, Max or Ultra subscriber.

Which one should you pay for?

All of this assumes you have decided to pay at all, and whether Claude Code earns back its price at your own hourly rate is the arithmetic behind that decision. If what you are weighing is the subscription itself rather than the terminal tool, whether Claude Pro is worth it for the way you actually use Claude starts from a different question: how often the free limit stops you in the first place.

• Occasional or hobby coding: Pro. One tool, one bill, enough for a few sessions a week. Details of the allowance are in the Pro plan price and its limits.
• Sustained daily coding: Max or Ultra. Max gives five times Pro usage in both windows; Ultra gives ten times Pro usage each week and removes the five-hour limit for verified accounts. The choice between Max and Ultra comes down to how often you hit the weekly limit and whether you need to work without a five-hour ceiling.
• Several people: Team or Enterprise seats, where the allowance follows the seat (or, on a current usage-based Enterprise plan, every request is billed at API rates under admin spend limits) and an admin can see spend per user. What a seat costs against what the same work costs per token is the comparison to run before you buy seats in bulk.
• Bursty or automated runs: API billing. A nightly job or a CI (continuous integration) step has no reason to compete with your interactive work for the same allowance.

Those last two cases mix well. Keep the subscription for the work you do by hand, and give automation its own API key, so an overnight run cannot empty the window you need the next morning. The API and subscription cost comparison works through where the crossover sits.

FAQ

Can I use Claude Code on the free Claude plan?

No. Claude Code requires a Pro, Max, Ultra, Team, Enterprise or Console account. Anthropic's setup documentation states that the free Claude plan does not include Claude Code access, so running claude will ask you to sign in to a paid account or to authenticate with an API key instead. The free plan still covers chat in the web and desktop apps, which is enough to test whether Claude answers your kind of question before you pay for anything.

Does Claude Code use the same usage limits as Claude chat?

Yes on Pro, Max and Ultra. Anthropic states that usage limits are shared across Claude and Claude Code, so all activity in both tools counts against the same limits. Pro and Max have a session limit that resets every five hours, and all three plans have a weekly limit that resets at a fixed time each week assigned to your account. Verified Ultra accounts have no five-hour limit; unverified accounts have a temporary limit. Max and Ultra cap Fable at half of the regular weekly allowance, drawing from that same budget and using it faster than other models. In practice an afternoon of heavy refactoring can leave you with a short chat window that evening. Run /usage to see the bars and the next reset time.

Why did my usage drop so fast during one coding session?

Because a coding turn carries much more context than a chat message. Every request sends the whole conversation, and each tool result adds another request that carries it again. Run /usage and read the behaviour flags: anything responsible for 10 percent or more of recent usage is marked, and the usual culprits are long context and cache misses. A session left open all day keeps re-sending its full history, and an idle gap longer than the cache lifetime (one hour for the main conversation on a subscription within plan usage, five minutes for subagents and once you are on usage credits) makes the next message write everything to the cache again at the cache-write rate, which is above the normal input rate: 2 times base input for a one-hour cache write and 1.25 times for a five-minute one. Using /clear between unrelated tasks addresses both.

When is an API key cheaper than a Pro or Max subscription?

When the work is bursty or automated. API billing charges per token with no monthly floor, so a script that runs twice a week costs what those tokens cost and nothing more, while a subscription pays for a month of capacity you never use. Sustained daily coding runs the other way: a subscription is a fixed price for a large allowance, and paying per token for that same volume costs more. Setting ANTHROPIC_API_KEY on a machine where you also code by hand is the mistake to avoid, because once you approve the key at the prompt, or run claude -p with it present, Claude Code bills the API and leaves your subscription allowance unused.