Claude Code usage limits: how the 5-hour and weekly limits work in the terminal

On Pro and Max plans, Claude Code draws on the same five-hour session limit and weekly limit as claude.ai. The Pro and Max article says “all activity in both tools counts against the same limits.” Team plans and seat-based Enterprise work the same way, and each request counts against both windows at once.

Anthropic publishes no number of hours, prompts or tokens for any plan, and the pricing page says “there’s no fixed message count”. The Free plan doesn’t include Claude Code.

With an API key, there are no plan limits. You pay per token, within the API’s rate limits and spend limits.

Which limits apply depends on how you signed in

Anthropic’s Claude Code limits article ties the meter to how you signed in. Find your row in the table.

Signed in with	What limits you	Source
Pro or Max	Session and weekly limits shared with chat and IDEs; Max adds a Fable cap of up to 50% of the weekly limit	Pro and Max article, Fable article
Team or seat-based Enterprise	A per-seat allowance on a rolling five-hour window and a weekly window, shared with chat and Cowork; its size depends on the Standard or Premium seat	Costs guide
Usage-based Enterprise	“no per-seat usage limits”; usage is billed at API rates	Team and Enterprise article
API key (Console, Bedrock, Vertex, Foundry)	Per-token billing, with rate limits and a monthly spend cap set by usage tier on the Console, or your cloud provider’s limits	Limits article, rate limits docs

claude -p and the Agent SDK draw on your plan limits when you’re signed in with a subscription. With an API key from a Console organization linked to a Max or Team plan, they spend that plan’s monthly API credits.

One catch from the Pro and Max article: if ANTHROPIC_API_KEY is set in your environment, Claude Code uses that key instead of your subscription, and you pay API charges.

One pool, two windows at the same time

In Claude Code’s errors reference, Anthropic’s example is a large workflow fanout: one burst “can exhaust the weekly allowance before the session window resets.” Once the weekly limit is spent, you wait for its reset, not the next five-hour window.

Switching models helps only with some limits. The errors reference says “the session and weekly limits are shared across all models, so switching models doesn’t restore access.” One older page disagrees: the limits article tells Enterprise seat users at a limit that they “can switch to a lighter model with /model”. Separate Opus and Sonnet limits apply only to their own model family, so switching to another family with /model keeps you working. Anthropic doesn’t say which plans have them.

If you work in both the terminal and the browser, the combined figure is what matters. Token Counter, our Chrome extension, shows session and weekly usage as percentages in a bar under the claude.ai message box and on its toolbar icon, with the time until each limit resets. Claude Code usage is already in those figures, because the extension reads the same claude.ai usage data as the Claude settings page. We suggest it for Pro and Max only, as it is not verified on the Free plan. Its only site access is claude.ai, so it doesn’t read your terminal or the Claude Code CLI. It shows plan totals. For a split by skill, subagent, plugin or MCP server, use /usage in Claude Code. It doesn’t show API-key, Console, Bedrock or Vertex usage, which isn’t plan usage. The extension needs you signed in to claude.ai in Chrome. It refreshes about once a minute while usage is changing and about every five minutes otherwise, so a sudden burst shows up late. It never pauses or stops Claude Code.

What uses up the limit fastest in Claude Code

Anthropic’s costs guide, limits article and errors reference name six main drains:

1. The conversation is resent every time. A one-line question late in an all-day session “still draws usage for the whole conversation.” Run /clear when you switch tasks (“the single most effective lever for both quality and cost”) and /compact when you are mid-task.
2. Cache misses. After a break longer than the cache lifetime (an hour on a subscription), your next message reprocesses your full context. Switching models does the same, because each model keeps its own prompt cache.
3. Subagents and agent teams. Each subagent sends its own requests on top of the main conversation, and agent teams “use approximately 7x more tokens than standard sessions when teammates run in plan mode”.
4. Opus and thinking. Opus “uses meaningfully more of your quota”. To spend less, /model opusplan uses Opus for planning and Sonnet for execution. Thinking tokens bill as output tokens and can’t be turned off on Opus 5.5, Sonnet 5.5, Haiku 5.5 or Fable. Lower /effort instead.
5. @ mentions. The @ prefix pulls in the whole file plus its CLAUDE.md tree. A bare path saves tokens.
6. Idle time. Goal check-ins and scheduled tasks send your full context while you’re away, and background jobs cost “typically under $0.04 per session”.

On a paid plan, /usage “flags behaviors that account for 10% or more of your recent usage”. The usage walkthrough covers the rest of that screen.

What Claude Code does when you hit a limit

Before the limit, you may get a warning like the first line below. At the limit, Claude Code blocks new requests until the reset time in the message. The errors reference lists the wording:

You've used 85% of your session limit · resets 3:45pm
You've hit your session limit · resets 3:45pm
You've hit your weekly limit · resets Mon 12:00am
You've hit your Opus limit · resets 3:45pm
You've hit your Sonnet limit · resets 3:45pm

If you hit the session limit mid-response, Claude Code may finish the current step. The wrap-up article says this wrap-up doesn’t apply with an API key, and the guide on when limits reset covers it.

Automatic continue. By default, Claude Code waits in the open session and picks the task up where it stopped once the limit resets. The interactive mode docs limit this to interactive sessions with a claude.ai login, on v2.1.234 or later.

• It doesn’t resend your last message.
• It still asks for permissions as usual, and it re-arms the wait at most twice in a row.
• It won’t start the wait on its own for a reset more than 24 hours away, as a weekly one can be. You can still start it from the usage-limit menu, /rate-limit-options. -p runs, background sessions and API keys get no wait.
• To turn it off, switch off “Continue automatically at usage limit” in /config, or set autoContinueAtUsageLimit to false.

Other options. Wait for the reset, run /usage-credits to continue at API rates (see the usage credits guide), or upgrade: the Pro vs Max comparison shows how much more each plan gives.

Errors that look like a limit but aren’t

These four are not plan limits:

• Server is temporarily limiting requests (not your usage limit): a short-lived throttle unrelated to your plan quota, retried automatically since v2.1.199.
• Request rejected (429): the rate limit set for your API key, Amazon Bedrock project or Google Cloud project. A stray ANTHROPIC_API_KEY can route you through a low-tier key. Check the active credential with /status.
• Usage credits required for 1M context: an entitlement check, not an exhausted quota.
• A full context window is different from a usage limit. /compact or /clear fixes it.

How Claude Code’s limits changed in 2026

As of October 2026:

• May 6, 2026. Anthropic’s May 6 announcement said it was “doubling Claude Code’s five-hour rate limits for Pro, Max, Team, and seat-based Enterprise plans” and removing the peak-hours reduction on Claude Code for Pro and Max. The help pages still describe one pool for chat and Claude Code and don’t explain how a doubled Claude Code limit fits inside it.
• June 15, 2026. The Agent SDK article says planned changes to Agent SDK usage were paused, so claude -p and the SDK still draw on your subscription limits.
• September 22, 2026. The Opus 5.5 announcement said Anthropic was “increasing five-hour usage limits on Pro, Max, Team, and seat-based Enterprise plans,” with no figures. It also gave subscription users a rate limit reset they can save and use whenever they choose.
• October 7, 2026. An update to the same Agent SDK article says Max and Team plans now include monthly API credits for claude -p and the SDK. Interactive Claude Code can’t use them.

Some sites still quote limits as hours per week. Those figures are not on Anthropic’s current plan pages.

Token Counter is an independent extension and is not affiliated with Anthropic.