How to ask your company for more Claude Code budget

1. Read the message to see whose limit it was. The table in the next section says what each message means and who can change it.
2. If it was your seat's 5-hour window, the cheapest thing is to wait: it resets in full five hours after it started, and on a subscription Claude Code continues the task by itself after the reset unless you choose otherwise in /rate-limit-options.
3. If you cannot wait, ask. On a Team or Enterprise plan, /usage-credits in Claude Code asks you to confirm and then sends the request to your admins; running it again does not send a duplicate. On a seat-based Enterprise plan, the Claude app also shows a "Request usage credits" link when you hit your limit. On a plan your company pays for some other way, such as a Max plan on a company card, ask the person who holds the card.
4. Before you ask, run /usage and note what it shows: both bars, the reset times, and the breakdown of what has been using them. That is the start of your evidence.

Claude Code's error reference lists the messages a member of a Team or Enterprise plan can see. On organisations billed by usage, the messages say "usage limit" where these say "spend limit".

If one of your seat's windows is also what ran out, the message adds when it resets, for example "your session limit resets 3:45pm", and access comes back then without anyone raising anything. On a Team plan, a Premium seat has "5x more usage than standard seats". If you hit the Standard seat's window every day, asking for a Premium seat may be cheaper for the company than paying for your usage credits, and it is a simpler request to approve. If you are not on a Claude Team or Enterprise plan, for example on a Max plan your company reimburses, there is no admin screen: the request goes to whoever approves the expense.

In public threads, the people on the other side of the request describe the same problem: they see a request for more and cannot tell what it is for. Three of them, quoted without names:

• A finance or IT person: "there are cases where engineers (in most of the cases but not only) reach their usage limits and ask for additional usage and we can't tell if it's justified and really contributes to the productivity or not".
• Another: "If people are hitting usage limits they need to make a business case why we need to raise their cap, not just swipe the company credit card with no oversight."
• An engineer describing their company: "they want to see cold hard numbers in how much it'll save in money, and no not even time savings count."

So the request that gets approved is the one that answers the approver's question before they ask it: what was the allowance spent on, and was it the work the company wanted done?

Keep it short. A request your manager can forward to their own manager without rewriting it is the one that moves. Five lines are enough:

1. What ran out, in the words of the message: your session or weekly limit, or a spend limit, and how often it has happened this month.
2. What you were working on when it did: the project, the ticket or the pull request, not "coding".
3. What filled it, if you know: long sessions, several agents running at once, a large codebase being read. /usage gives you the breakdown for the last 24 hours or 7 days on the machine you run it on.
4. What you need: a Premium seat, usage credits with a monthly limit of a stated amount, or a higher cap, and for how long.
5. What happens without it: the work waits for the reset, or moves to a tool the company has not approved.

If your company asks for a project code or a written justification, as some do, the same five lines fill it in.

Of the 138 people we found on Reddit who said they ran out on an account their employer pays for, 29 described asking a manager, IT or a department head for more, which was the most common thing they said they did. Some describe one request; others describe a ladder:

• "Now as the new sprint started, I needed more limit to start implementing. Was asked by my manager to approach his boss with everything, as to why I need the said limit."
• "Can make further requests (up to +$200 per request) providing project codes...", followed in the same comment by three levels of approval up to the head of development.
• On GitHub Copilot rather than Claude, the same shape: "We get $120 per month and have to ask an officer for more than that.. each month... With justification."
• And the cost of asking often: "I don't want to talk to my manager every few days to ask him to extend my limits."

That last one is the reason to ask once, with evidence, for a change that lasts the month, rather than for a top-up each time.

/usage shows the last 24 hours or 7 days on one machine, and Settings > Usage on claude.ai shows where you are now. Neither keeps a record of the times you ran out or what you were doing. OpenReason Intel, a dashboard for Claude Code and Codex usage, does: your 5-hour and weekly windows over time from every machine you use, each time a window filled, and the sessions and turns that were running, each with its estimated share of the window. That is the "how often" and "on what" of the request, in one place you can open on screen with your manager or invite them to.

It does not add to your allowance, and it does not tell you anything before you run out. It is a record you look at afterwards, and a record you can show. The share of the window it gives each turn is an estimate, and it says so.

Anthropic's admin screens show each member's month-to-date usage-credit spend and a spend report a day behind, but not each member's 5-hour or weekly window or what filled it. The guides on who went past their plan and what your team spends on Claude Code cover what you can see and set today, and what happens when your team's Claude credits run out covers the shared-budget case. OpenReason Intel shows every person's windows and the turns around each over-limit period, for Claude Code and Codex, so a request can be read against the record.