Your ChatGPT Workspace Is Out of Credits — and the Two Switches That Fix It Both Ship With No Cap

Key takeaways

• Running out of credits is a block, not a bill. OpenAI's Business help page puts it plainly: if the workspace doesn't have enough credits for a usage-based feature, "that feature may be unavailable until credits are added." Nothing is accruing while you're stopped.
• The fix has a mirror image. Both of the settings that keep you from being stopped again — automatic reload on Business, the workspace overage limit on Enterprise/Edu — arrive uncapped by default.
• "No limit" means two opposite things depending on which screen you're looking at. OpenAI flags this itself: on a group override it means inherit the workspace default; on the workspace overage limit it means no cap at all.
• A blank field is a decision. On Business, the Monthly recharge limit box left empty "allows unlimited automatic reload purchases each month."
• Alerts are not brakes. "Alerts do not block usage. Use the workspace overage limit when you need a cap."
• Zero is a real setting, and it's the one that behaves the way most founders assume the default already does.
• Your credits have a shelf life — 12 months from purchase — which makes "buy a big block so this stops happening" a worse idea than it looks.
• No employees, no Codex seat, no workspace? Then this is a five-minute read for later. The pattern — a meter with the cap switched off until you switch it on — is the part worth remembering.

The message is a stop sign, not a charge

If you're reading this because a workspace told you it was out of credits, start with the reassuring half.

OpenAI's Business help page describes the behavior in one sentence: "If your workspace does not have enough credits for a usage-based feature, that feature may be unavailable until credits are added." That is a block. The feature stops. You are not quietly running a tab.

That distinction matters because the instinct in the moment is to make the block go away as fast as possible — and the fastest routes out are also the two settings that decide how large your bill can get. This brief is about the second half of that trade, which the error message doesn't mention.

Two things worth separating before anything else:

These are different products with different switches. ChatGPT Business manages this through credits you buy and an automatic reload setting. ChatGPT Enterprise and Edu manage it through a committed credit pool plus a workspace overage limit and per-user usage limits. The vocabulary overlaps enough to be confusing and the screens are not the same. Everything below is labeled by plan, because applying the Enterprise answer to a Business workspace will send you looking for a setting that isn't there.

This brief isn't about what a credit costs. We worked through the rate side separately — which runs are metered and how to estimate a month — in ChatGPT workspace agents pricing: what credits will actually cost you. Here the question is narrower: what happens when they run out, and what the two ways of preventing that actually commit you to.

On Business: the blank field that means "unlimited"

Adding credits on Business is four steps, and the fourth carries a fact that doesn't appear in the error message: "Review the payment method, credit amount, and note that credits are valid for 12 months."

Hold that thought, because it changes the calculus on the obvious move — buying one large block so the interruption never recurs. Credits bought against a burst of usage that doesn't repeat are credits with an expiry date on them.

The setting that actually prevents interruptions is automatic reload. It has three fields, and the third is where the money is:

"You can set a Minimum balance so ChatGPT automatically charges your payment method on file to add credits when your balance goes below that amount and tops up to your Target balance. … To cap automatic reload purchases, enter a Monthly recharge limit; leaving this field blank allows unlimited automatic reload purchases each month."

Read the last clause the way a founder should read it: the field that caps your monthly exposure is optional, and leaving it alone is not neutral. It's the setting that permits the largest number.

There's a second-order surprise in the same paragraph worth knowing before you flip the switch, not after:

"If your workspace balance is already below the Minimum balance when you turn on automatic reload, ChatGPT may charge the payment method on file right away to add credits, up to the Target balance and subject to any Monthly recharge limit."

So the natural sequence — you get blocked, you go turn on automatic reload to stop it happening again — is precisely the sequence that can trigger an immediate charge, because you are by definition below your minimum at that moment. That isn't a trap so much as an ordering problem: set the Monthly recharge limit in the same sitting, not as a follow-up task.

And on the per-seat side, Business spend controls start open too: "By default, all seats and users have no limits specified." You can set a monthly credit ceiling by seat type or by individual user, and a user-level value overrides a seat-level one — but until you do, there is no ceiling in place.

One aside that will save somebody a support ticket: if your plan is Business and you're hunting for Codex seats you don't have, the page carries a note that "Starting June 24, 2026, Codex seats will no longer be available to new ChatGPT Business plans, or to Business workspaces that never added a Codex seat prior to June 24 2026." Workspaces that had one before that date can keep managing and adding them. If you're new and looking for a door that people on forums describe walking through, that door closed.

On Enterprise and Edu: "No limit" means two opposite things

Here is the part that is genuinely difficult, and to OpenAI's credit the help page flags it in its own voice rather than leaving you to discover it:

"Note: Both Usage limits and the Workspace overage limit have a 'No limit' setting option. 'No limit' has different meanings depending on the configuration surface."

The two meanings:

"For a group override set in Usage limits, No limit means the group inherits the workspace default; it does not mean unlimited usage."

"For the workspace overage limit, configured in the global admin console, No limit means no cap is configured on eligible overage usage. No limit does not make credits free or unlimited; eligible overage usage can continue and may be billed according to your agreement."

Same two words. On one screen they mean fall back to the house rule; on another they mean there is no house rule. An admin who has learned the first meaning — the reassuring one — and then meets the same label on the billing surface has every reason to read it the same way. It doesn't read the same way.

And the second surface is the one that ships that way out of the box:

"OpenAI does not set a workspace overage limit by default. The setting appears as No limit until a workspace owner defines one. If no overage limit is set, certain feature usage can continue being utilized uninterrupted, but could incur unwanted overage charges."

Note who can act on it. Usage limits are configurable by "workspace admins and owners"; the overage limit lives in the Global Admin Console; and "Only workspace owners can view invoices." The person who first notices the spend is not necessarily the person who can see the bill or set the cap.

Two layers, and they stop different people

Usage limits and the overage limit are often discussed as if they were the same guardrail. They stop different things:

The order of resolution for user limits is stated explicitly — "user override, highest applicable group default, then workspace default" — and if someone belongs to two groups, the higher limit wins unless they have an individual override. That is worth knowing before you assume a restrictive group assignment is doing restrictive work.

The three sentences that decide whether you get a surprise

Three lines on the Enterprise/Edu page do more practical work than the rest of it combined.

One: alerts don't stop anything.

"Usage alerts notify selected recipients when the workspace reaches a chosen credit threshold, uses all committed credits, or reaches the overage limit. Alerts do not block usage. Use the workspace overage limit when you need a cap."

If your entire control setup is an alert, you have monitoring, not a limit. The page says so in the imperative.

Two: zero is a real value, and it's the one that matches most people's mental model of a default.

"Setting the limit to 0 prevents the workspace from continuing into overages. Some advanced features may pause until additional credits are available or the limit changes."

That second sentence is the honest trade, and it is the same trade as the message that sent you here — a feature pauses. The difference is that you chose it, at a moment when you weren't mid-task.

Three: a cap is not a hard stop at the exact number.

"The credit value for some usage cannot be determined exactly until a task completes. A task that starts under the limit may finish and push usage slightly over the limit. Once the limit is reached, usage controls prevent additional eligible usage."

This is a reasonable engineering consequence of billing work that hasn't finished yet, and it's the kind of detail worth reading before you set a limit at exactly the number you cannot exceed.

One more, for anyone doing the arithmetic after the fact: "Changing the overage limit affects future eligible usage. It does not retroactively remove usage that already occurred or change an invoice that has already been issued." And when you go looking for that invoice, the period may not be the month you're picturing — overage invoices "follow the workspace's contract billing period, which may be anchored to the workspace's subscription start date rather than the calendar month."

What to actually do, by plan

If you're on Business:

1. Add credits to clear the block — and note the 12-month validity while you're on that screen, which argues for topping up to what you'll plausibly use rather than to a round number that feels safe.
2. Turn on automatic reload and fill in the Monthly recharge limit in the same sitting. Blank means unlimited; a number means a number.
3. Expect a possible immediate charge when you enable it if you're already below your Minimum balance — which you are, if you just got blocked.
4. Set per-seat or per-user credit maximums. They start unset.

If you're on Enterprise or Edu:

1. Open the Global Admin Console and look at the workspace overage limit. If it says No limit, that is not a fallback — it's an absence.
2. Decide deliberately between a number and 0. Zero reproduces the behavior you're used to (things pause) with none of the surprise.
3. Set usage limits for the people whose work you'd actually want throttled, remembering that the highest applicable group default wins.
4. Add alerts on top — but as instrumentation, not as the cap.

If you have no workspace yet: nothing here is a task. It's a pattern to recognize. Metered features increasingly arrive with the meter running and the cap switched off, and the setting that protects you is the one you have to go find.

When your tools grow meters, keep something flat

The reason this is a founder problem and not an IT problem is that it's rarely one tool. A stack assembled over a year quietly converts from flat seats to metered usage, one product at a time, and each conversion ships with its own screen, its own vocabulary, and its own idea of what "no limit" means. The work isn't learning any single billing page; it's noticing that your fixed costs have become variable while you weren't looking.

Which is an argument for keeping the boring layers boring. The parts of a company that don't need to be metered — where your mail arrives, where filings reach you — are worth having at a price you can state from memory. For the US side of that, our partner save office handles addresses and mail; Auteur doesn't operate the US service directly, and there's an overview on our US virtual office page.

FAQ

What does "Your workspace is out of credits" actually mean? That a usage-based feature has stopped because the workspace's credit balance is insufficient. OpenAI's Business help page states it as: "If your workspace does not have enough credits for a usage-based feature, that feature may be unavailable until credits are added." It describes a block on the feature, not a charge that keeps accruing. Adding credits — or, on Enterprise/Edu, having room under the workspace overage limit — is what restores it.

Who can add credits or raise the limit? On Business, adding credits requires signing in "with a role that is allowed to manage workspace billing or credits," then going to Workspace settings → Billing. On Enterprise/Edu, "workspace admins and owners can manage usage limits in Workspace settings," admins and owners configure alerts and the overage limit in the Global Admin Console, and "Only workspace owners can view invoices." These can be three different people in the same company.

Do ChatGPT workspace credits expire? On Business, yes. The step for adding credits says to "note that credits are valid for 12 months." That's worth weighing against the instinct to buy a large block after an interruption.

Will setting an overage limit stop my team mid-task? Setting it to 0 "prevents the workspace from continuing into overages," and OpenAI notes that "some advanced features may pause until additional credits are available or the limit changes." A non-zero limit isn't an exact stop either: "a task that starts under the limit may finish and push usage slightly over the limit." Whichever value you pick, the behavior at the boundary is a pause, not a failure.

Are usage alerts enough? No, and the documentation is unusually direct about it: "Alerts do not block usage. Use the workspace overage limit when you need a cap." Alerts tell you what happened; the overage limit is what decides whether it can keep happening.

I can't find Codex seats on my Business plan. As of the note on OpenAI's Business credits page, "Starting June 24, 2026, Codex seats will no longer be available to new ChatGPT Business plans, or to Business workspaces that never added a Codex seat prior to June 24 2026." Workspaces that added one before that date "can continue to manage and add usage-based Codex seats." If you're setting up a new Business workspace now, that route isn't available regardless of credits.

Bottom line

The message that stopped you is the safe state. A feature is unavailable until credits are added — nothing is running up in the background while you decide.

What deserves the same minute of attention is the shape of the exit. On Business, the Monthly recharge limit is an optional field, and leaving it blank "allows unlimited automatic reload purchases each month." On Enterprise and Edu, OpenAI "does not set a workspace overage limit by default," and the words that describe that absence — No limit — mean inherit the default one screen over. Both defaults resolve toward more spend, not less, and both are one field away from not doing that.

Set the number in the same sitting you clear the block. That's the whole recommendation.

This is general information about a vendor's published billing settings as of July 31, 2026, not financial or contractual advice. OpenAI's help pages change; confirm the current behavior on the pages linked above, and read your own agreement for how overage usage is billed.