Claude Code Limits: What Each Plan Actually Gets
Claude Code does not have a message count, and looking for one is why the limits feel unpredictable. What it has is a rolling five-hour session window, a weekly window on top of that, and a single pool shared with everything else you do on your Claude account.
That last part surprises people most. Your work in the terminal, your chats on claude.ai, the desktop app, and mobile all draw from the same allowance. A morning spent in Claude chat is a morning already spent in Claude Code.
The windows
| Window | What it does |
|---|---|
| Rolling 5-hour session | The primary limit on every plan, free and paid |
| Weekly | Added on paid plans, sits on top of the session window |
| Daily routine runs | A separate cap on how many routine runs can start per account |
Anthropic is explicit that there is no fixed number behind these: how much you can do “depends on the length and complexity of your conversations, the model you choose, and the features you use.” It also reserves the right to limit usage “in other ways, such as weekly and monthly caps or model and feature usage.”
Relative sizing is the only published shape. Pro gives at least 5x more usage per five-hour session than Free. Max gives 5x or 20x more than Pro. On Team plans, Standard seats give more than Pro and Premium seats give 5x more than Standard.
Claude Code itself is included in all paid plans, so the tier you are on is the limit you have. See Claude Code pricing for what each one costs.
Reading your own numbers
Run /usage. On a paid plan it shows plan usage bars, activity stats, and a breakdown of what has been consuming them:
- Attribution: recent usage split across skills, subagents, plugins, and individual MCP servers, each as a percentage. An MCP server is charged only for requests that actually consumed one of its tool results.
- Behavior flags: things like long context or cache misses, flagged once one accounts for 10% or more of recent usage.
Press d or w to switch between the last 24 hours and the last 7 days. Two caveats worth knowing before you trust the number: the figures come from local session history on that machine, so usage from another laptop or from claude.ai is not included, and the Session block’s dollar figure is priced at list rates for API users rather than reflecting what a subscriber is billed.
If the usage endpoint is rate limited, /usage falls back to the last bars it loaded within the past hour and says so. Press r to retry.
The six things that actually burn the allowance
Long sessions consume far more than the amount of typing in them suggests. The documented causes:
Long context. Claude Code sends your whole conversation with every request, and each tool call sends another request carrying those results. A one-line question in a session that has been open all day still draws on the entire history.
Cache misses. Your first message after a gap longer than the cache lifetime reprocesses the full context. The lifetime is an hour on a subscription and drops to five minutes once you are drawing on usage credits.
Scheduled tasks. A scheduled task fires on its interval even while the session sits idle, sending the full context each time.
Cross-session messages and goal check-ins. Both start a new turn in an idle session, each carrying the full context again.
Agent teammates. Each active teammate runs its own context window and keeps consuming until it exits. Agent teams use roughly 7x the tokens of a standard session when teammates run in plan mode, which is why they are off by default.
Compaction. /compact has to read the conversation it summarizes, so compacting a large context is itself a large request. /clear costs nothing.
The single highest-leverage habit is clearing between unrelated tasks, and the second is matching the model to the job rather than leaving Opus as the default.
Four messages that look alike and are not
| What you see | What it means |
|---|---|
| ”You’ve hit your session limit” | The rolling five-hour window. Switching models does not restore it |
| ”You’ve hit your weekly limit” | The weekly window. Same: model switching does not help |
| ”You’ve hit your Opus limit” | Model-specific. /model to Sonnet keeps you working |
| Auto-compact warning | Not a limit at all. The conversation is near its compaction threshold |
The first two are shared across all models, which is the part people get wrong when they try to escape a limit by switching.
When you hit one
Three options: wait for the window to reset, move to a higher plan, or turn on usage credits and keep working on metered overage.
/usage-credits opens the right place depending on who you are. Pro and Max subscribers land on their claude.ai usage settings, where credits can be switched on and the balance, month’s spend, and monthly spend limit are visible. Team and Enterprise members with billing access get the organization’s usage settings. Members without billing access send a request to their admins instead, and running the command again while one is pending tells you so rather than sending a duplicate.
On Pro and Max, hitting your spend limit with credits still available prompts you to raise or remove the limit without leaving the CLI.
Routines have their own ceiling
Routines run in Anthropic’s cloud, so they keep working with your laptop closed, and they draw on subscription usage exactly like an interactive session. On top of that they have a daily cap on runs started per account. One-off scheduled runs do not count against that cap, though they still draw normal usage. During the research preview, GitHub webhook events also face per-routine and per-account hourly caps, and events beyond the limit are dropped until the window resets.
Worth knowing before you plan around them: every routine requires one or more GitHub repositories, cloned on each run, and the trigger vocabulary is a schedule with a one-hour minimum interval, a GitHub pull request or release event, or an HTTP call to a per-routine endpoint you wire up yourself.
What teams should budget
On Team and Enterprise plans each member’s usage draws from a per-seat allowance on the same rolling five-hour and weekly windows, shared with Claude chat and Cowork, sized by seat tier.
For API-billed organizations, Anthropic’s published planning figures are roughly $13 per developer per active day and $150 to $250 per developer per month, with 90% of users staying under $30 per active day. Budget more for a coding seat than a chat seat, because one debugging session carries file contents, tool calls, and multi-step reasoning.
Per-user recommended rate limits scale down as the team grows, from 200k to 300k tokens per minute per user at 1 to 5 users, to 10k to 15k at 500 or more, because fewer people run concurrently in larger organizations.
The usage that should never have been on this meter
Look at the list of things that burn the allowance and one pattern stands out: the expensive usage is the recurring usage. The scheduled task that fires whether or not you are there. The routine that runs nightly. The long-lived session you never cleared.
That is the work least suited to a coding plan’s meter, because it is the same shape every time. Triage the inbox each morning, chase the quote that went quiet, file the receipt, update the record when the deal moves: none of that needs a model reasoning from scratch on every run, and none of it should be competing with your debugging session for the same five-hour window.
Carly is built for that half. Its Zapier-style workflow steps are free, so the route-and-file-and-notify work that runs hundreds of times a month is not the thing you are rationing. Carly’s triggers are keyed per mailbox and per calendar, so a workflow can fire when mail lands in the client account and stay silent on the personal one, with nothing open and no repository involved. AI agents start at $35/month.
And because Carly runs an MCP server, the two do not compete. Register it once and Claude Code can call Carly’s email, calendar, CRM, and workflow tools directly, which means those actions happen through Carly rather than against your Claude allowance:
claude mcp add --transport http --scope user carly https://carlyassistant.com/mcp/FAQ
How many messages do you get with Claude Code?
There is no message count. Usage is measured against a rolling five-hour window plus a weekly window on paid plans, and how far it stretches depends on conversation length, model choice, and which features you use.
Does Claude Code share limits with Claude chat?
Yes. Activity on the web, desktop, mobile, and Claude Code all draws from the same pool. On Team and Enterprise plans the seat allowance is shared with Claude chat and Cowork too.
Will switching models get me past a limit?
Only for a model-specific one. An Opus limit is escaped by switching to Sonnet with /model. Session and weekly limits are shared across all models, so switching changes nothing.
Why did a short session use so much?
Almost always long context or a cache miss. Claude Code resends the whole conversation on every request, and the first message after a break longer than the cache lifetime reprocesses all of it. /clear between unrelated tasks is the fix.
Is Claude Code included on the free plan?
Claude Code is included in all paid plans. See Claude Code pricing for the tiers.
Related: Claude Code pricing · What is Claude Code · Multiple Gmail accounts in Claude Code · Claude Code alternatives · Cursor pricing · Codex pricing
Ready to automate your busywork?
Carly schedules, researches, and briefs you—so you can focus on what matters.
See what people say
"Before Carly, I relied on a Calendly link, but the whole process felt impersonal and not very professional. Carly changed that by handling all the back-and-forth, so I'm no longer stuck in endless email threads trying to line up schedules.
Now Carly reaches out to candidates, shares my real-time availability, lets them pick a slot, then sends a Zoom link and drops it straight into my calendar. She sends reminders to both of us before each call, which has significantly reduced no-shows and last-minute confusion.
On top of scheduling, Carly acts like a full executive assistant, sending me my schedule the night before so I can prepare for each call. It reminds me of the old x.ai assistant, but Carly is noticeably smarter, faster, and better suited to my healthcare recruitment business."
