LIMITS

Claude Code token limits

Identify context pressure, plan allowance and API restrictions in Claude Code. Find the right commands and next steps.

GUIDESources checked
On this page 6 sections

Find the kind of limit first

Claude Code can encounter context pressure, a product usage allowance, or API billing and rate restrictions. The message you see matters more than a generic token number. Save the exact message and check how you authenticated before choosing a response.

Use the displayed symptom
What you seeInspectNext action
Context is filling/contextSummarize or start a fresh task
Plan limit or reset message/usage and account usageFollow the displayed reset and allowance
Spend or billing restrictionAuthenticated billing accountCheck account budget and billing
API request-rate errorAPI limits and response headersFollow retry guidance

Inspect the context

Use /context to see what is loaded. Use /compact when continuing a long task; use /clear when moving to a new task whose old conversation is no longer useful. They manage conversation context. An account usage reset is a separate event.

Before clearing a useful session, preserve the decisions and unfinished work in your own notes. A concise handoff can help the next conversation continue without carrying every earlier tool result. This is a workflow suggestion, not a promise of additional account allowance.

Check account usage and billing

Current Claude Code documentation uses /usage for usage information. The available blocks depend on your setup. Verify the active account and model with /status and /model. Older versions and some documentation use /cost for session costs; consult your installed command help.

An API dollar estimate does not tell you the remaining percentage of a subscription. Follow the account's displayed reset and spend information. See Claude Code token usage for a step-by-step check.

Why counting one prompt is insufficient

A coding task can involve multiple model requests. Files, tool definitions, results and conversation history can be included in those requests. The short instruction you typed is only one part of the work.

Use the local counter to inspect a text sample, then observe real session usage for the complete workflow. When planning API spending, enter observed input, output and cache quantities in the Claude calculator.

A practical troubleshooting sequence

  1. Read the displayed message and note whether it mentions context, account usage, spend or request rate.
  2. Inspect the active account, model and relevant usage view.
  3. Address that category: shorten context, wait for the shown reset, adjust an authorized spend setting, or follow API retry guidance.
  4. Retry a small task and check whether the same restriction remains.

Avoid changing several settings at once. A small, repeatable task makes it easier to determine which change resolved the problem.

Keep future sessions measurable

Record a few representative tasks and inspect their usage after completion. Separate a quick edit, a repository investigation and a long automation run. Compare like tasks with the same model and similar project material. Use those observations to decide whether to change your workflow or account setup.

Sources & review

Checked 3 October 2026. Product rules can change; consult the linked official references.