How to Check Your Claude Token Usage (Every Method, Every Plan)

Last updated 19 September 2026. Menu labels, commands, meters, and API headers were checked against Anthropic’s current Help Center and product documentation. Interfaces change, so use the meter inside your own account as the final authority. [VERIFY: 2026-09-19]

Not affiliated with Anthropic. Claude Token Optimization is an independent site and audit service. Claude® and related product names are trademarks of Anthropic PBC. Plan prices, model rates, and usage limits change; always confirm on Anthropic’s own docs. [VERIFY: 2026-09-19]

If you are searching how to check Claude usage limit, the confusing part is that there is no single universal counter. Pro and Max show capacity bars. Claude Code has commands inside the terminal. API responses expose rate-limit headers, while the Console reports organization-level traffic and billing. Each answers a different question. [VERIFY: 2026-09-19]

Start by identifying what stopped you. A subscription session bar, a weekly limit, a full context window, an API 429 response, and a spending threshold are not the same problem. Read why you keep hitting Claude’s usage limit for the diagnosis, and how Claude tokens work for the unit underneath it.

Method map: where every Claude usage meter lives

Method 1: Claude.ai Settings > Usage

For Pro, Max, and other eligible seat plans, begin in the web app:

1. Sign in at claude.ai.
2. Select your profile initials.
3. Open Settings.
4. Select Usage, then review Plan usage limits. [VERIFY: 2026-09-19]

The Current session bar represents the rolling session allowance, commonly framed around a five-hour window. The Weekly area can show separate treatment for Opus and all other models where Anthropic exposes that split. The displayed reset time is more reliable than calculating it from memory. [VERIFY: 2026-09-19]

Usage credits, when enabled and available for the plan, let work continue after included capacity is consumed. That continuation is metered separately rather than enlarging the included bar. [VERIFY: 2026-09-19] Claude across the web, desktop, mobile, and Claude Code can draw from a shared subscription pool, so a quiet web afternoon may still follow a heavy coding morning. [VERIFY: 2026-09-19]

There is still no official hard token allotment shown for every Pro subscriber. See how many tokens Claude Pro gives you for the honest capacity explanation, and when Claude usage resets for both clocks.

Method 2: Claude Code commands

In Claude Code, enter /usage. On a subscription session, this opens the current usage display so you can inspect session and weekly capacity without leaving the terminal. Press d or w to switch attribution between the last 24 hours and the last 7 days; press r to retry when the usage request fails or shows last-known data. [VERIFY: 2026-09-19]

1. Open Claude Code in the terminal for the account you want to inspect.
2. Type /usage and press Enter.
3. Confirm whether Claude Code is authenticated to a subscription or an API key.
4. In VS Code, open Account & usage for the extension’s view. [VERIFY: 2026-09-19]

Authentication matters. With subscription access, the useful question is how much included capacity remains. With an API or Console key, /cost reports estimated API cost for the session; it does not turn API billing into a Pro capacity bar. [VERIFY: 2026-09-19] If the current client exposes /usage-credits, use it to inspect or manage paid continuation separately from included plan usage. [VERIFY: 2026-09-19]

Method 3: API headers and the Console

API users should inspect each response rather than waiting for an error. Anthropic returns rate-limit headers that describe request and token constraints. Depending on the endpoint and account, look for families including: [VERIFY: 2026-09-19]

• anthropic-ratelimit-requests-* for request capacity and resets;
• anthropic-ratelimit-tokens-* for combined token controls;
• anthropic-ratelimit-input-tokens-* for input-specific controls;
• anthropic-ratelimit-output-tokens-* for output-specific controls; and
• retry-after when the service tells the client how long to wait. [VERIFY: 2026-09-19]

Log the limit, remaining, and reset values your response actually contains. A 429 means the current rate boundary was exceeded; it does not automatically mean the account’s monthly budget is gone. Back off using the returned timing instead of retrying aggressively.

In the Console, use Usage for measured traffic, Rate limitsfor account constraints, and Settings > Billing for balances and spend controls. [VERIFY: 2026-09-19] Those views answer organization questions that a single local Claude Code session cannot. To turn token volume into a planning estimate, use the Claude token cost calculator.

Method 4: add the habit layer the product meter cannot show

A bar tells you how much remains, not which behavior consumed it. Log the date, surface, model, approximate task size, turns, and whether you hit a limit in the usage tracker. After a week, patterns such as long file threads, Opus by default, or parallel coding sessions become easier to see.

Teams need the same habit layer across people because organization totals can hide one expensive workflow. The team usage tracking guidecovers that roll-up without pretending the product provides a perfect per-task ledger.

Which limit did you actually hit?

After you can see the meter, change the work

Measurement matters only if it changes the next session. Start a new conversation when the task changes. Stop re-attaching durable files. Route routine work away from the heaviest model. Disable tools you are not using, and reduce parallel agent loops. The planned guide to cutting Claude token usage turns those ideas into a checklist.

If the meter still drops faster than the visible work suggests, the missing cost is often in system instructions, repeated context, tool definitions, or conversation history. The The audit findings guide explains those patterns. You can also have your own prompts reviewed, or read the audit example before deciding. No account or call is needed.

FAQ

How do I check my Claude usage limit?

On claude.ai, open your profile initials, then Settings > Usage and review Plan usage limits. In Claude Code on a subscription, run /usage. [VERIFY: 2026-09-19]

Can I see exactly how many Claude tokens I have left?

Subscription plans generally expose capacity bars and reset timing rather than one universal exact token balance. API users can measure billed tokens and inspect rate-limit headers. [VERIFY: 2026-09-19]

How do I check Claude Code usage?

Run /usage. If Claude Code uses an API key, use /cost for the local session estimate and the Console for organization usage and billing. [VERIFY: 2026-09-19]

Where can I see API rate limits?

Inspect the anthropic-ratelimit-* headers returned with API responses and check the Console’s Rate limits area. Respect retry-after on limited responses. [VERIFY: 2026-09-19]

Why does Settings > Usage disagree with my message count?

Messages have different weight. Files, conversation history, model choice, tools, and work on other surfaces in the shared pool can move capacity faster than a visible message count. [VERIFY: 2026-09-19]

Next steps

• Diagnose the meter: why you keep hitting the limit.
• Track the habit: log a week of real sessions.
• Estimate API traffic: use the cost calculator.
• Find hidden waste: have your prompts reviewed or see an example.