Why does my Claude Code 5-hour limit run out so fast?
by vortwangUpdated
Usually because every request re-sends your whole conversation. Anthropic's cost docs list the causes: 'Claude Code sends your full conversation with every request', a cache miss after a break longer than 'an hour on a subscription' reprocesses the full context, scheduled tasks and cross-session messages fire while idle, agent teams use 'approximately 7x more tokens', and /compact is itself a large request. Run /usage: on Pro, Max, Team and Enterprise it flags whichever behavior accounts for 10% or more.
What the official docs say
Claude Code sends your full conversation with every request
Quoted from Claude Code docs: Manage costs effectively — Why usage climbs in a long session The lifetime is an hour on a subscription and drops to five minutes once you're drawing on usage credits
Quoted from Claude Code docs: Manage costs effectively — Why usage climbs in a long session Agent teams use approximately 7x more tokens than standard sessions when teammates run in plan mode
Quoted from Claude Code docs: Manage costs effectively — Manage agent team costs (approximately 7x more tokens) Thinking tokens are billed as output tokens, and the default budget can be tens of thousands of tokens per request
Quoted from Claude Code docs: Manage costs effectively — Adjust extended thinking (/effort, MAX_THINKING_TOKENS) flagged when one accounts for 10% or more of recent usage
Quoted from Claude Code docs: Manage costs effectively — Plan usage breakdown in /usage some Claude Code sessions to spawn excessive parallel subagents, burning through usage faster than expected
Quoted from @ClaudeDevs on X, 2026-06-01: limits reset after a bug spawned excessive parallel subagents, "burning through usage faster than expected"
How it is counted
Start from /usage. On a Pro, Max, Team or Enterprise plan it shows 'a breakdown of what counts against your plan limits': attribution to skills, subagents, plugins and MCP servers, plus behavior flags 'such as long context or cache misses, flagged when one accounts for 10% or more of recent usage'. That tells you which row of the table below applies before you change anything.
The biggest lever is context. The docs say 'a one-line question in a session that has been open all day still draws usage for the whole conversation'. /clear starts a fresh session at no cost; /context shows what fills the window; /compact summarizes it but 'compacting a large context is itself a large request'.
Cache lifetime matters on a subscription. 'Your first message after a break longer than the cache lifetime misses the cache and reprocesses your full context.' The lifetime 'is an hour on a subscription and drops to five minutes once you're drawing on usage credits'. Anthropic publishes no absolute message count for the five-hour window; the Max plan article defines it only relative to Pro.
Users report fast drains (for example issue #90675: Max 5x, five-hour limit exhausted about 23 minutes after a reset). Those are user reports, not a published rate. Anthropic did confirm one systemic cause on 2026-06-01: a bug that made 'some Claude Code sessions spawn excessive parallel subagents, burning through usage faster than expected', fixed with a reset for everyone.
| Cause | What the docs say | What to do | Source |
|---|---|---|---|
| Long context | Full conversation sent with every request; each tool batch is another request carrying it | /context to see what fills the window; /clear for a fresh start (costs nothing); /compact to summarize | [1] |
| Cache miss after a break | First message after a break longer than the cache lifetime reprocesses the full context; lifetime is an hour on a subscription, five minutes on usage credits | Resume from a summary when offered; keep long sessions small before stepping away | [1] |
| Scheduled tasks (/loop, /schedule) | A scheduled task fires on its interval even while idle, sending your full context each time | /usage shows a Loops row per task; stop tasks you no longer need | [1] [4] |
| Cross-session messages | Delivered as a new turn while the session sits idle, sending your full context each time | Set crossSessionInbound to hold | [1] |
| Agent teams and subagents | Approximately 7x more tokens when teammates run in plan mode; each teammate keeps consuming until it exits | Keep teams small, use Sonnet for teammates, shut teammates down when done | [2] |
| Extended thinking | Thinking tokens billed as output; default budget can be tens of thousands per request | /effort to lower the level; MAX_THINKING_TOKENS on fixed-budget models | [3] [5] |
| Compaction | /compact reads the conversation it summarizes, so compacting a large context is itself a large request | Prefer /clear when you do not need continuity | [1] |
| Published messages per 5 hours | Not published; Max 5x and 20x are defined as five and twenty times the Pro per-session allowance | Watch rate_limits.five_hour.used_percentage in the status line | [6] [9] |
FAQ
Is there a published number of messages per five hours?
No. Anthropic defines Max 5x and Max 20x only relative to Pro (five and twenty times the per-session allowance) and says usage varies with message length, conversation length and model. Reports such as "gone in 23 minutes" are user reports, not a published rate.
Which command shows what drained my window?
/usage. On Pro, Max, Team and Enterprise it attributes recent usage to skills, subagents, plugins and MCP servers, flags long context or cache misses when they account for 10% or more, and lists the heaviest scheduled tasks. Press d or w to switch between the last day and week.
Does stepping away for lunch cost anything?
The next message can. On a subscription the prompt cache lasts an hour; after a longer break the first message reprocesses the full context. On Pro and Max, resuming a large session after a long break offers to resume from a summary so later requests do not carry the full history.
Do subagents and agent teams count against my 5-hour limit?
Yes. Each teammate runs its own context window, and the docs put agent teams at approximately 7x more tokens than a standard session when teammates run in plan mode. On 2026-06-01 Anthropic reset everyone after a bug spawned excessive parallel subagents.
Sources
- Claude Code docs: Manage costs effectively — Why usage climbs in a long sessionverified 2026-09-18
- Claude Code docs: Manage costs effectively — Manage agent team costs (approximately 7x more tokens)verified 2026-09-18
- Claude Code docs: Manage costs effectively — Adjust extended thinking (/effort, MAX_THINKING_TOKENS)verified 2026-09-18
- Claude Code docs: Manage costs effectively — Plan usage breakdown in /usageverified 2026-09-18
- Claude Code docs: Slash commands — full command table (/usage, /usage-credits, /rate-limit-options, /status, /cost, /stats)verified 2026-09-18
- Claude Help Center: What is the Max plan? (session resets every five hours; weekly at a fixed time assigned to your account)verified 2026-09-18
- anthropics/claude-code issue #90675 (2026-08-30): Max 5x, entire 5-hour usage limit exhausted ~23 minutes after resetverified 2026-09-18
- @ClaudeDevs on X, 2026-06-01: limits reset after a bug spawned excessive parallel subagents, "burning through usage faster than expected"verified 2026-09-18
- Claude Code docs: Status line — Rate limit usage (rate_limits.five_hour / seven_day)verified 2026-09-18
Want the next Claude reset on your phone?