When Claude Code says a usage limit has been reached, keep the exact message, run /usage, and check /status before changing models, accounts, or billing. If another screen appears to show remaining usage, first verify the active Claude identity, organization, API credential, balance type, and named limit.
A shared session or weekly allowance, a model-family limit, a context warning, an API 429, and a spend limit require different fixes. Use only a path Anthropic exposes for the authorized account: wait for the displayed reset, enable eligible extra usage, ask an organization administrator for capacity, change plan for recurring demand, or intentionally use a separately billed Console API account.
Identity and spend-limit checks reviewed October 12, 2026. Error text, commands, models, limits, reset rules, Usage Credits, and rates can change. Treat the live message,
/usage,/status, account billing page, and Anthropic documentation as the source of truth.
Have remaining usage but still see a limit? Check this first
Start with this identity and routing check:
- Claude identity: Are you signed in with the email and sign-in method that owns the expected plan?
- Organization: Is Claude Code attached to the intended personal, Team, or Enterprise organization?
- Credential override: Does
/statusshow a subscription identity or an API route? Check whetherANTHROPIC_API_KEY, a cloud-provider credential, or a gateway setting is taking precedence. - Balance type: Are you viewing included plan usage, Usage Credits, or Claude Console API balance? They are not interchangeable.
- Named meter: Does
/usageidentify a session, weekly, model-family, or extended-context allowance? - Other Claude activity: Does usage from Claude chat, Cowork, or another supported surface share the applicable allowance?
- Service or API condition: Is the message actually a rate limit, spend limit, authentication error, gateway failure, or service incident?
Remaining personal capacity does not rule out an organization cap. The official cost guide notes that the Team/Enterprise usage-credits row does not show an organization-wide limit. It also describes a Showing last-known usage notice when plan usage cannot refresh: check the timestamp and use the offered refresh action before treating those bars as current.
If the numbers still appear wrong, save the full message, timestamp and timezone, client version, model, active organization, authentication method, and a redacted screenshot before contacting Anthropic Support. Never include a key.
Classify the exact Claude Code limit
The official Claude Code error reference and usage-and-limits guide separate these conditions. Clearing context cannot refill plan usage, and waiting for a plan reset cannot fix an API project's spend cap.
When the message names an organization spend limit
A search phrase such as org spend limit reached is not enough to identify who owns the limit. Preserve the full error, then use /status to check the active account and authentication route. Claude Code's authentication reference explains how a configured API credential can be active instead of a subscription login.
Anthropic's organization cost guide separates these routes. A subscription organization's usage-credit cap is not a Console API cap. Ask the owner to review the named budget or wait for the applicable reset; increasing a personal plan or buying unrelated credits may not change it.
For the direct API, the rate-limit reference distinguishes:
- Monthly tier cap: HTTP 429,
enforced_spend_limit_reached, and noretry-after; repeated retries cannot restore access. - User-configured budget: commonly HTTP 400, with a message naming the limit and restoration time.
- Claude Code workspace exception: its cap can return 429 with
retry-after. That header alone does not prove a short-lived request-rate limit.
Use the response details and owning billing page together. Only an authorized owner should approve a budget change; do not remove controls or substitute someone else's key. After the stated reset or an approved change, try one small request on the intended route before resuming the interrupted task.
If the shared session or weekly allowance is exhausted
Use the exact reset shown by /usage or the error. Switching between Claude models does not restore a fully exhausted shared window.
Supported next steps can include:
- waiting for the displayed reset;
- enabling eligible Usage Credits with a spending cap;
- asking a Team or Enterprise administrator to review capacity controls;
- changing plan when representative work repeatedly reaches the correct limit;
- deliberately using a separate Console API route with its own budget and policies.
The Claude Code pricing guide compares those commercial paths; the Claude Code workflow guide explains the separate Workspace decision. Do not submit the same prompt repeatedly while waiting; inspect files and partial output first.
If only a named model family is limited
A model-family limit can be narrower than the shared session and weekly windows. If /usage shows that shared usage remains and the exact notice identifies one family, /model may let you choose an available model outside that family.
Confirm that the new model is authorized and suitable for the task. Start with a bounded, reversible step and re-run the relevant test. Do not treat model switching as a workaround when the message names the shared allowance.
Usage Credits and Console API billing are different
Anthropic's Usage Credits guidance describes eligible extra usage attached to a paid Claude plan. Availability, balance, standard API-rate charging, and spending controls are account-dependent.
A Claude Console API key is a separate credential and bill. Before changing routes:
- run
/statusand identify the active method; - check environment and managed settings for credential overrides;
- confirm who owns the Console organization and may spend from it;
- set a budget and rate limit;
- store the key in an approved secret manager;
- understand that plan Usage Credits do not fund the Console API balance.
Do not borrow another person's key or share a login. That changes authorization and risk; it does not solve the account's legitimate capacity problem.
Context limits and service errors need different fixes
/compact summarizes conversation history so the same task can continue with less context. /clear removes conversation history and is appropriate only after required decisions and handoff details are preserved. Neither command replenishes plan usage.
For an API 429, first identify whether the request uses Anthropic, a cloud provider, or a gateway. Inspect the error details for a spend cap before applying that route's retry policy. For a spend-limit message, inspect the billing owner for the active route. For a transient service incident, preserve work and retry conservatively instead of running a tight loop.
Protect work before retrying
Before another request:
- review saved files and the current diff;
- preserve test output and unresolved decisions;
- identify any external action that may already have run;
- record the last accepted step;
- retry only the missing, bounded action.
After capacity returns, reduce waste by limiting irrelevant context, pointing to exact files, keeping instructions focused, disabling unused integrations, and checking partial work before retrying. Usage depends on model, context, tools, parallel sessions, and repair loops—not a universal prompt count.
What Agent.Space changes—and what it does not
Agent.Space can preserve project files, Sessions, and reviewed handoffs between supported Agents. The guide to how Agent.Space works explains that Workspace continuity separately from Anthropic usage.
It does not refill a Claude allowance, grant an extended-context entitlement, authorize a borrowed key, or bypass Anthropic account rules. Choosing Agent.Space is a separate product and billing decision.
Compare current Agent.Space plans, then create an Agent.Space account if the displayed model access, paid option, and Workspace workflow fit your needs.