Claude Code usage limit by plan, and what 17 of its limit and error messages mean

What are the Claude Code usage limits on each plan?

Claude Code's usage limit on every subscription is a slice of one account wide budget, metered per five hour session and per week, and Anthropic sizes each plan only as a multiple of Pro. No plan page prints a prompt or token count. Anthropic's pricing FAQ says "there's no fixed message count", so the roughly 45 prompts per five hours on Pro that guides still list is an outside estimate, not Anthropic's number. Every cell below comes from Anthropic's pricing page, its plan help articles and its API rate limits reference, read on 18 October 2026.

Divide price by multiple and the ladder flattens. On monthly billing, Pro, Max 5x, a Team Standard seat and a Team Premium seat all cost $20 per Pro sized session. Max 20x is the one step that halves it, to $10. Team limits are per person, too: Anthropic's Team plan page says "Usage limits on Team plans are per-member, rather than applied to the team as a whole."

Two rows have no window at all. Usage based Enterprise bills by consumption at API rates, and a Console API key has "No hard stop; the account is charged for what it uses", per Anthropic's Claude Code help page. The key still has ceilings, just different ones. Requests over a tier's per minute limits get a 429, and an organization that spends its tier's monthly cap ($500 on Start, $1,000 on Build, $200,000 on Scale) is paused until 00:00 UTC on the first of the next month. Fable is the other exception. On Pro it runs only on usage credits, and on Max it gets its own weekly bar.

What does Claude Code say when you hit a limit?

Claude Code names the meter that ran out in the first words of the message, so read those before you do anything else. "Session" and "weekly" mean your plan's windows. A model name means a limit on one family. "Spend limit" means a usage credits cap, and "429" means an API key. We compiled every limit message from Anthropic's Claude Code error reference, interactive mode docs and its Pro and Max help page on 18 October 2026. The times shown are Anthropic's examples; yours will differ.

The first two rows are the ones switching models can't fix. Anthropic's cost docs say a session or weekly limit is "shared across all models, so the developer can't restore access by switching models with /model." The Opus and Sonnet rows are the opposite: a model outside that family keeps you working right away. The last two rows aren't usage limits at all. A 529 "doesn't count against your quota", and if Claude itself is struggling, is Claude down shows it.

When does the Claude Code usage limit reset?

Claude Code's session limit resets every five hours on a rolling window, and its weekly limit resets at a fixed time each week that Anthropic assigns to your account. Neither one resets at midnight. The limit message prints the reset time, and so does /usage.

Since version 2.1.234, Claude Code can wait for you. When a claude.ai limit stops a task mid run, it holds the open session and shows "Continuing automatically at 3:45pm · esc to cancel". At the reset it prompts Claude to pick the task up where it stopped, without resending your last message. It re-arms the wait at most twice in a row. Then it stops with "Automatic continue stopped after repeated usage-limit hits". Three cases skip the automatic wait:

• A weekly reset more than 24 hours away. Claude Code opens /rate-limit-options instead, and you pick "Wait here, then continue automatically" yourself.
• An Opus or Sonnet limit while you're on another family. Your next turn may never hit it, so there's nothing to wait for.
• API keys, cloud providers and usage based billing. Anthropic says usage there "is metered per request, so there is no reset to wait for."

One early refill exists. Anthropic hands out a free limit reset now and then, and the Opus 5.5 launch on 22 September 2026 gave subscribers one "which you can now save and use whenever you choose." You can't spend it from the terminal: Anthropic's help page says "The "Reset for free" button isn't currently available on Claude Mobile or in Claude Code in your terminal or IDE." Use it in Settings > Usage on the web or in Claude Desktop, and Claude Code gets the full limit too, because the budget is shared. An API organization on a spend cap has no such button. It waits for the calendar month to turn, at 00:00 UTC.

How do you check Claude Code usage with /usage and /status?

Run /usage to see your plan's session and weekly bars with their reset times, and run /status to see which account and credential Claude Code is actually using. Those two commands answer most "why am I blocked" questions between them. /cost and /stats are aliases for /usage.

On Pro, Max, Team and Enterprise, /usage also breaks down what spent the allowance: skills, subagents, plugins and each MCP server as a share of the total, plus flags for any behavior, such as long context or cache misses, that took 10% or more. Press d or w to switch between the last 24 hours and the last 7 days. Read that toggle with care. Anthropic says those figures are "computed from local session history on this machine, so usage from other devices or claude.ai is not included." The plan bars are the account wide number, and the breakdown is only this laptop.

/status is the check for a wall that makes no sense. If it shows an API key row, Claude Code is billing that key, not your plan. Anthropic's Pro and Max help page warns that with ANTHROPIC_API_KEY set, "Claude Code will use this API key for authentication instead of your Claude subscription." That's how a Max subscriber ends up with a 429 or "Credit balance is too low". Unset the variable, relaunch, and run /login. To watch the meter all day, a custom status line can read rate_limits.five_hour.used_percentage and rate_limits.seven_day.used_percentage. Anthropic says those fields appear only for Pro and Max subscribers, or behind a company gateway that sets a spend limit.

How have Claude Code usage limits changed since 2025?

Anthropic has raised Claude Code's five hour limit faster than its weekly cap. It doubled the five hour limits on 6 May 2026 and raised them again on 22 September, while the weekly limit ended September only 25% above its April level. Every row below is from Anthropic's own announcement or help page, read on 18 October 2026, except the first, which is Anthropic's email as TechCrunch reported it.

Put the two multipliers side by side and the week got tighter in session terms. If one five hour window holds at least twice what it did in April and the week holds 1.25 times as much, a week now absorbs at most 62.5% as many maxed out sessions as it did then. A developer who maxes out every session now reaches the weekly wall in fewer of them than in April. That's our arithmetic on Anthropic's figures, not a number Anthropic states, and the 22 September raise makes the true share lower still.

The Opus weekly cap is history. Anthropic's Opus 4.5 post said "we've removed Opus-specific caps", and the plan pages now describe one weekly limit "across all models". Claude Code still carries the "You've hit your Opus limit" message, though, and Anthropic's docs keep the /model fix for it. Fable is the model with its own meter now: on Max it can use "50% of weekly limits", per the pricing page.

Why does Claude Code hit the limit so fast?

Claude Code sends your whole conversation with every request, and every tool call adds another request carrying the results. Anthropic's cost docs put the consequence in one line: "a one-line question in a session that has been open all day still draws usage for the whole conversation." Five habits drain the allowance fastest:

• Long sessions. Context grows each turn, so the hundredth prompt costs far more than the first. /clear between unrelated tasks "costs nothing", while /compact is itself a large request.
• Breaks past the cache lifetime. The prompt cache lasts an hour on a subscription and drops to five minutes once you're on usage credits. Your first message after a longer break reprocesses everything.
• Opus as the default. Anthropic says Sonnet "handles most coding tasks well and costs less than Opus", and a switch to Opus also moves the subagents that inherit your model.
• Subagents, agent teams and loops. Each spawns its own requests. Agent teams use "approximately 7x more tokens than standard sessions" when teammates run in plan mode, and a /loop task fires on its interval even while you're idle.
• Everything else on the account. Chats on claude.ai, Claude Desktop, scripts run with claude -p and the Agent SDK all draw on the same limits.

Two cheap fixes cover most of it. Keep CLAUDE.md under about 200 lines, since it loads at the start of every session, and move long workflow instructions into skills that load only when used. Then run /mcp and switch off servers you aren't using. The /usage breakdown will tell you which of these is your problem, because it flags any behavior that took 10% or more of recent usage.

How do you keep working when Claude Code's limit is reached?

The fastest route depends on the message. A model family limit clears with /model in seconds. A session or weekly limit leaves four choices, and they get more expensive in this order:

1. Let it wait. Keep the session open and Claude Code continues the task after the reset on its own, or pick "Wait here, then continue automatically" from /rate-limit-options for a weekly reset.
2. Turn on usage credits. Run /usage-credits. On Pro and Max it opens Settings > Usage, where credits are prepaid and "billed at standard API rates". Usage bundles cut that: $50 of usage costs $45, $250 costs $200 and $1,000 costs $700, up to $2,000 a month. On Team and Enterprise, a member without billing access sends the admin a request instead.
3. Switch to a Console API key. You pay per token with no five hour window, at $4 in and $20 out per million for Opus 5.5 and $2 and $10 for Sonnet 5.5. The monthly API credits that come with Max "can't be used in Claude Code or the Claude apps."
4. Upgrade. Run /upgrade. Max 20x is the only step that lowers the price per Pro sized session, to $10.

Use the API figures before you pick between options 3 and 4. Anthropic says Claude Code across enterprise deployments averages about $13 per developer per active day and $150 to $250 per developer per month, with 90% of users under $30 per active day. That range sits either side of the $200 price of Max 20x, and the API has no weekly cap. If you hit the weekly wall every week, the API or Max 20x is the fix. If you hit it once a month, credits are cheaper. Is Claude Max worth it does the full plan math.

Claude Code doesn't run on Whizi, which publishes this page: Anthropic lists a Claude plan, a Console API key, Amazon Bedrock, Google Cloud and Microsoft Foundry as the ways to pay for it. What it can do is answer the coding question in chat while your window refills. Whizi Pro is $29.99 a month for 2,500 credits, and Claude Sonnet costs 10 credits a message and Claude Opus 20, so that's 250 Sonnet or 125 Opus messages a month with no five hour window (Claude in Whizi). Where it loses: credits refill only when the UTC month turns, and every plan stops at 60 messages an hour.