Updated guide · June 2026
Claude Code does not have a separate quota. On Pro and Max it draws from the same usage pool as Claude.ai chat and Cowork, under the same rolling five-hour session limit and weekly limits. Because agent work does so much per task, it can drain that shared pool far faster than chatting.
Straight answer
Claude Code has no separate quota — it draws from the SAME usage pool as Claude.ai chat and Cowork. Rough estimates: ~10–45 prompts per 5-hour window on Pro, far more on Max (up to ~900 on Max 20x), with weekly Sonnet hours ~40–80 (Pro) to ~240–480 (Max 20x). Agentic work (file reads, tool calls, long context) burns more. Verify in Claude → Settings → Usage; numbers change.
5-hour rolling window plus the weekly schedule. Verify in Claude → Settings → Usage; numbers change.
Shared pool
The single most important fact about Claude Code limits is that there is no separate Code budget on Pro or Max. Chat, Cowork and Code all spend from the same pool, so a heavy coding session can leave little left for chatting and vice versa.
Claude.ai chat, Cowork and Claude Code all draw on the same plan usage. Spending in one reduces the others within the same five-hour and weekly windows.
Pro users can run Claude Code, not just Max. But because it shares Pro's pool, agent work is often the real reason Pro feels tight.
On May 6, 2026 Anthropic doubled the five-hour limits and removed peak-hour reductions, and weekly limits have been temporarily raised at times. That gives Code more room, but the figures still change.
There is no published number of Code prompts. It depends on model, task size, files touched, tool calls and capacity, so always read Settings then Usage rather than trusting a quota table.
Estimates
Anthropic does not publish a prompt count for Code. The numbers below are rough third-party estimates only, useful for rough planning, not promises. Your real limits live in Settings then Usage.
| Plan | Rough est. prompts / 5h (unofficial) | Why it varies | Practical note |
|---|---|---|---|
| Pro ($20/mo) | ~10 to 45 prompts per 5-hour window, very approximate | Model choice, task size, files read, tool calls and current capacity all move this widely. | Fine for light coding; heavy agent runs can exhaust it quickly. Save Opus for hard tasks. |
| Max 5x ($100/mo) | Roughly five times Pro per session, unofficial | Same variables as Pro, but a larger shared pool to spend from. | Comfortable for regular daily coding without constant interruptions. |
| Max 20x ($200/mo) | Much more than Pro, roughly twenty times per session, unofficial | Same variables; the most headroom for long agent sessions. | Best fit for all-day Code and Cowork use. Still not unlimited. |
Treat every figure here as a rough estimate that changes. The only authoritative view of your remaining usage is Claude under Settings then Usage.
Why it burns
A single Claude Code task is not one message. It is a chain of reads, reasoning, edits and commands, and each step spends from your pool. That is why one coding request can cost as much as many chat messages.
Understanding a repo means opening and reasoning over lots of files, each adding context and cost before any code is written.
Searching, editing, running tests and commands are separate steps. A task with dozens of tool calls spends far more than a one-shot answer.
Agent sessions accumulate large context as they work, and that context is reprocessed on each turn, raising cost over a long task.
Running Code on Opus consumes far more of your pool than Sonnet. Use Sonnet for routine edits and reserve Opus for genuinely hard problems.
Give focused, well-defined tasks. Vague or sprawling requests cause more exploration, more tool calls and more usage.
Check Settings then Usage before a big run so you know how much shared session and weekly headroom you have for both Code and chat.
Windows
Claude Code obeys the same two layers as chat, but spends them faster. Knowing which limit you are hitting tells you whether to pause briefly or rethink the week.
Starts with your first message and resets about five hours later. Heavy agent runs can empty this window quickly, so plan large tasks in deliberate bursts.
The Pro and Max weekly caps apply to Code too, with separate reset info for Opus versus other models. If you hit these, waiting hours will not help; lighten model use or reduce scope.
Because chat, Cowork and Code share one pool, a big coding day reduces your chat budget for that window. Sequence heavy work so it does not starve the surface you need next.
For the underlying plan mechanics, see our Claude Pro limits and Claude Max limits guides, and always confirm the live numbers in Settings then Usage.
Shared pool scenario
The shared pool is easiest to grasp through a single day. Because Claude Code, Claude.ai chat and Cowork all draw on the same five-hour and weekly budget, work in one quietly shrinks what is left for the others. Here is how that plays out, illustratively, on Pro.
Your first Claude Code message opens the five-hour window. The task reads many files on Opus, makes dozens of tool calls and runs tests. By 10:00 it is done, but it has spent a large share of the window.
You move to Claude.ai to draft documentation. It is the same pool, so the budget you have left is what the morning Code run did not use. A few long Opus chats here add up faster than you expect.
You ask Cowork to work across some files. This also spends from the shared pool. Around now Claude slows or pauses you, even though no single surface felt heavy on its own.
Settings then Usage shows the five-hour session is nearly spent. A short wait will reset it. If instead the weekly cap were the binding one, waiting would not help and you would lighten model use or reduce scope.
The lesson is to sequence heavy work. If you know you need chat headroom this afternoon, do not spend the whole morning on Opus agent runs in the same window.
Window mechanics
The rolling five-hour session limit is the layer Claude Code users hit most, because agent work spends it fast. Understanding exactly when it starts and resets removes most of the surprise.
| Time | What happens | Window state |
|---|---|---|
| 10:00 | Your first message of the day starts the session window. | Window opens, full budget available, resets around 15:00. |
| 10:00 to 12:00 | A heavy Opus Code run reads files, makes many tool calls and edits. | Large share of the budget spent in two hours. |
| 13:00 | More chat and a second agent task. | Remaining budget runs low; Claude may slow or pause you. |
| 15:00 | Roughly five hours after the first message. | Session resets; the next message opens a fresh window. |
Times here are illustrative. The exact reset is shown in Settings then Usage, and remember the weekly limits sit above this window and do not reset every five hours.
Reduce burn
You can get far more done in the same window by changing how you drive the agent. The goal is less wasted exploration and smaller context, since that is what costs the most.
Give focused, well-defined instructions. Vague or sprawling requests cause more file reading, more tool calls and more usage as the agent hunts for what you meant.
Use /compact to trim a long Claude Code session so its accumulated context stops being reprocessed in full on every turn. Long agent sessions get more expensive the longer they run.
Point Claude Code at only the files and directories that matter for the task rather than the whole repository. Less context per turn means lower cost per turn.
Default to Sonnet for routine edits, refactors and tests, and reserve Opus for genuinely hard reasoning. Running everything on Opus is the fastest way to drain the shared pool.
Plan choice
Because Claude Code shares your plan pool, the plan you are on is the real ceiling on agent work. This table maps coding intensity to the right option, including when to step outside subscriptions entirely.
| Coding pattern | Best option | Why |
|---|---|---|
| Light, occasional coding | Pro ($20/mo) | Enough for small tasks; heavy runs can exhaust the shared pool quickly. Save Opus for hard problems. |
| Regular daily coding | Max 5x ($100/mo) | A larger shared pool that handles steady daily agent work without constant interruptions. |
| All-day, Opus-heavy agent work | Max 20x ($200/mo) | The most headroom for long sessions, still not unlimited, with the two weekly caps in play. |
| Automated or bursty programmatic use | Anthropic API | Pay per token with no five-hour or weekly caps; best for pipelines and integrations you control. |
Pricing labels above are approximate and change, so verify on the official pricing page. Many heavy developers keep a plan for interactive Claude Code and the API for automation. For the underlying mechanics see our Claude Pro limits and Claude Max limits guides.
Alternatives
When agent work drains your shared pool, you do not have to stop. The most reliable approach is to not depend on one app's cap. MultipleChat lets you run Claude alongside other models so one provider's limit does not block you.
Run Claude-style answers next to ChatGPT, Gemini and search models in one workspace. When Claude Code drains your pool, keep reasoning and drafting in another model instead of waiting for a reset.
AlternativeBroad assistant with strong coding help when Claude is unavailable.
AlternativeUseful for Google Workspace and Google-context work.
ResearchResearch-focused answers with web sources for docs and APIs.
Our guideGuide to Claude alternatives by use case.
ComparisonDetailed comparison of Claude and ChatGPT for coding.
GuideThe same kind of limit guide for ChatGPT users.
Sources
Limits and behavior change often for agent tools. Use these official pages and your own Settings then Usage screen.
Open Claude and check Settings then Usage for your live shared-pool limits.
Official plan names, current pricing and usage wording.
Anthropic explains how Code consumes usage and the shared pool.
How Code shares your Pro or Max usage.
Details on what consumes usage and the length limits.
FAQ
Short, hedged answers for people searching Claude Code limit, Claude Code shared usage, Code 5-hour window and prompts per session.
Yes. On Pro and Max, Claude Code shares the same usage pool as Claude.ai chat and Cowork. Using one reduces what is left for the others. The same rolling five-hour session limit and weekly limits apply. Verify on the official support article.
Yes. There is no separate Claude Code quota on Pro or Max. Chat, Cowork and Claude Code all draw from one pool, so a heavy coding session leaves less for chatting in the same five-hour and weekly windows, and a chat-heavy morning leaves less for agent work. Settings then Usage shows the shared total.
Agent work reads many files, reasons across a codebase, edits, runs commands and makes many tool calls in a single task. Each step consumes usage, so one coding task can cost far more than one chat message, especially on Opus.
Every file Claude Code opens adds its contents to the working context, and every search, edit, test run and command is a separate step that consumes usage. A task that touches dozens of files and makes dozens of tool calls can cost as much as many chat messages, because all of that activity sits inside one request.
There is no published fixed number. Rough, unofficial third-party estimates suggest roughly 10 to 45 prompts per five-hour window on Pro and much more on Max 20x, but this varies widely by model, task size and capacity. Treat these as estimates only and check Settings then Usage.
The rolling five-hour session limit starts with your first message and resets about five hours later, the same as chat. Because Claude Code can spend that budget quickly, the window can run out faster during heavy agent work.
The five-hour session limit resets about five hours after your first message, so if only that is capping you a short wait usually clears it. The weekly limits reset on their own schedule, with separate timing for Opus versus other models, so if those are the bottleneck waiting hours will not help. Settings then Usage shows both.
Scope tasks tightly, keep the context small, use /compact to trim long sessions, point Claude Code at only the files that matter, and default to Sonnet rather than Opus for routine edits. Vague or sprawling requests cause more exploration and more tool calls, which is what spends the budget.
Yes. The weekly limits on Pro and Max apply to Claude Code usage too, since it draws from the shared pool, with separate reset info for Opus versus other models. See Settings then Usage for your current standing.
Pro can run Claude Code and is fine for light coding, but heavy agent runs can exhaust it quickly because the pool is shared. Max 5x suits regular daily coding, and Max 20x is built for all-day agent sessions on Opus. If Claude Code is your main use, the plan choice usually comes down to how many hours a day you run agent work.
Use the API for automated, programmatic or bursty workloads where you want no five-hour or weekly caps and pay per token. Use Claude Code on a Pro or Max plan for interactive, hands-on development where a flat price and the integrated tooling matter more. Many heavy users keep both, plan for daily coding and API for automation.
Yes. Pro can use Claude Code. On May 6, 2026 Anthropic doubled the five-hour limits and removed peak-hour reductions, which helps, but heavy agent work can still exhaust Pro quickly. Verify current behavior on the official pages.