Usage limit reached (Claude, Codex, Cursor): what to do
A usage limit message means your subscription plan's allowance for a time window is used up. It is not an API rate limit. Wait for the reset time shown, check your usage, buy extra usage credits, upgrade your plan, or switch to an API key. Claude and ChatGPT both document these options.
A usage limit is not a rate limit
A rate limit (HTTP 429) is about speed on an API key. A usage limit is about total consumption on a subscription plan over a window of time. The message looks like "You've hit your session limit · resets 3:45pm" or "You've hit your weekly limit" in Claude Code. Requests are blocked until the reset time in the message.
On Claude plans, usage is a budget shared across claude.ai, Claude Code and Claude Desktop. Anthropic lists what drains it: long conversations, the model you pick, extended thinking, web search, and connected tools. For team plans, Anthropic describes the allowance as resetting on a rolling five-hour window plus a weekly window, shared across models. A model-specific message, such as an Opus limit, applies only to that model family.
Your options when you hit it
- Wait. The reset time is in the message. Claude Code can also wait and continue the interrupted task after the reset on recent versions.
- Check where it went. In Claude Code run /usage to see plan limits, reset times and an attribution breakdown by subagents, skills and MCP servers.
- Buy extra usage. On Claude Pro and Max, /usage-credits opens the page to turn on usage credits. ChatGPT Plus and Pro can buy credits for Codex.
- Switch model family. If only the Opus or Sonnet limit is hit, /model to another family keeps you working. Session and weekly limits cover all models.
- Use an API key. Codex can run extra tasks through an API key at standard API rates, and Claude Code works with API billing too.
- Upgrade the plan for higher base limits.
Why you hit it faster than expected
Claude Code sends the whole conversation with every request, so one short question in a day-long session still costs the whole history. A first message after a long break can miss the prompt cache and reprocess everything. Subagents, scheduled tasks and idle check-ins each send their own requests. Anthropic's own advice is to clear between unrelated tasks, pick a smaller model for simple work, turn off MCP servers you are not using, and keep CLAUDE.md short.
Codex limits depend on the plan, the model, and how large and complex each task is. Tasks on the same model can use very different amounts.
Plan ahead so you do not get stopped mid-task
Do the heavy work early in a window, not at the end. Split big jobs into steps that each finish cleanly, and write down where you stopped, so that after a reset a fresh session can continue from the notes without rereading everything. Keep an API key or credits ready as a fallback if a deadline matters.
If several agents share one plan, one busy agent can use up the allowance for all of them, so watch /usage during long parallel runs.
Where a node helps, and where it does not
A node does not add allowance to your plan. It can make the allowance go further in two ways. Agents that share a thread and memory stop repeating each other's reading, and each fresh session can start from stored notes instead of a long replayed history. A node can also share part of an AI plan's tokens by link with someone you choose. Check your provider's terms before sharing anything tied to a personal account.
Do it with a NodePress node
- Create a node and name it for the project.
- Paste the one line into each agent so they share one thread and memory.
- Ask agents to store decisions in memory, then start new sessions often to keep history short.
- If a teammate has spare allowance, create a share link for tokens you are both allowed to use.
If you are simply over what your plan allows, a node will not change that. Waiting, credits, a higher plan or API billing are the real fixes.
Questions people ask
What does "claude usage limit reached" mean?
Your plan's allowance for the current window is used up. The message shows when it resets, and you can wait, buy usage credits, or upgrade.
How often do Claude usage limits reset?
Anthropic describes a rolling five-hour window and a weekly window for team plans. The exact reset time is shown in the error message.
Can I keep working when Claude says limit reached?
Yes, with usage credits, a different model family if only Opus or Sonnet is capped, or an API key. Session and weekly limits cover every model.
How do I check Claude Code usage?
Run /usage. It shows plan limits, reset times, and which skills, subagents and MCP servers used the most.
What happens when I reach my Codex limit?
You can wait, buy credits on ChatGPT Plus or Pro, or run extra tasks through an API key at standard API rates.
Related
- 429 too many requests / rate limit reached: what to do
- What to do when you run out of AI tokens mid-project
- How to reduce AI agent token costs
- Subscription or API key: which is cheaper for coding agents?
- How to get AI agents talking to each other
- How to get AI agents talking to each other