How to set Cursor and GitHub Copilot spend limits.

01Where to set spend limits for Cursor, Copilot and Claude Code.

Each coding tool keeps its limits in its own admin screens, with its own scope and its own behaviour at the limit. The table shows where each cap lives and what happens when a developer reaches it.

TakeawayFind out which plan each tool is on before you set anything, because the plan decides which screen holds the limit.

02How to set a Cursor team spend limit.

Cursor's help centre now describes two plan families: The X Subscription, managed at grok.com/dashboard, and legacy Cursor plans, managed in the Cursor dashboard. The team limit sits in a different place on each.

On The X Subscription.

1. Go to grok.com/dashboard/billing as a team admin.
2. Under Additional usage credits, find Monthly additional usage limit and click Edit.
3. Enter one amount for the whole team, either fixed or per seat, and save. Left unset, there is no cap.

The limit counts paid additional usage, which on this plan includes third-party models such as Claude or GPT, billed from prepaid credit. Buying credit, by hand or through Auto Top-up, neither counts toward the limit nor raises it. Only team admins can change the limit or the top-up settings.

On a legacy Cursor Teams plan.

1. Open the Spending tab in the Cursor dashboard.
2. Turn on on-demand usage, which has to be enabled before limits appear.
3. Set the monthly spend limit for the team. Changes take effect immediately.

On Cursor Enterprise.

Enterprise adds limits for members and groups. Set the team default in the Spending tab, a member override in the Members tab, and a group override in the Groups tab, or manage all of them through the Admin API. When a user has limits from several sources, Cursor applies the highest one. Changing the team limit does not update existing overrides, so remove an override to move someone back to the team default. Dynamic Spend Limits, a toggle in the Spending tab, scales the team limit with the number of seats.

At any limit, paid usage stops for that user only, the user sees a notification, and usage resumes at the start of the next period or when the limit is raised. Cursor says enforcement is not instant, so usage can briefly pass the limit.

TakeawaySet the team limit first, then add member and group overrides only where someone needs a different number.

03How the Cursor on-demand spend limit works.

On legacy Cursor plans, on-demand usage is pay-as-you-go usage beyond the plan's included pools, and the on-demand spend limit caps it each month. Setting the limit to No Limit removes the cap.

• Overage before enforcement. Usage that slips past the limit before Cursor enforces it is held as a temporary spend-limit credit. You are billed up to the current limit.
• Raising the limit mid-cycle. A higher limit in the same cycle can make that credited overage billable, up to the new limit.
• Who can change it. The Only Admins Can Edit Usage Settings toggle, under Permissions in the Spending tab, decides this. While it is off, any team member can turn on-demand usage on or off and change the team limit.
• Spend alerts. On a legacy Teams plan, click Add Alert in the team dashboard, choose team-level or member-level, and set a threshold. On The X Subscription, Teams admins set alerts under Billing, then Alerts. Alerts send an email and leave usage running.

TakeawayTurn on Only Admins Can Edit Usage Settings before you rely on the on-demand limit.

04How to set a GitHub Copilot budget for AI credits.

Copilot Business and Copilot Enterprise now meter usage in AI credits, where one credit is worth one US cent. Each licence adds a monthly allowance to a shared pool, 1,900 credits for Business and 3,900 for Enterprise, and metered charges begin once the pool runs out. Budgets are set in US dollars.

1. Go to your enterprise or organisation, open Billing and licensing, then Budgets and alerts.
2. Click New budget and choose Bundled AI credits budget to cover Copilot, Copilot cloud agent, Spark and other AI-credit features together.
3. Set a universal user-level budget: under Budget scope choose Users and leave the user field empty, so it applies to every Copilot-licensed user.
4. Add a user-level budget for a cost centre where one department needs a different per-person amount.
5. Create the enterprise spending limit and any cost centre or organisation budgets, and select Stop usage when budget limit is reached on each one.
6. Select Receive budget threshold alerts to get an email and a banner at 75, 90 and 100 percent.

Enterprise owners and billing managers can set every budget. Organisation owners can set organisation budgets only, and those can restrict usage further below an enterprise budget without overriding it. Whichever budget has the least headroom left blocks the user first, so a low enterprise limit can stop someone with plenty left in their own budget.

The enterprise budget caps metered charges and leaves licence fees outside it. When a user is blocked, Copilot features that use AI credits stop, while inline code completions and next edit suggestions keep working because they use no credits.

TakeawaySet a universal user-level budget before rollout, and turn on Stop usage when budget limit is reached on every spending budget.

05What happened to the Copilot premium requests budget.

Premium requests are now legacy billing in GitHub's documentation. GitHub converted existing premium request budgets into AI credit budgets automatically, and premium request billing now applies only to Copilot Pro and Pro+ subscribers on an existing annual plan who stayed on it after 1 June 2026.

For an organisation or enterprise, the old premium requests budget is now an AI credit budget on the same Budgets and alerts page. Review the converted amounts once, because a credit is priced by the tokens a request uses, so the same work can cost a different amount than it did under per-request billing.

TakeawaySearch your billing settings for AI credit budgets, since premium request budgets have been converted.

06How to control Copilot spend with policies by organisation and enterprise.

Policies decide whether Copilot can spend past the included pool at all, and which models developers can choose. Budgets decide how much.

• The AI credits paid usage policy. It is on by default for Copilot Business and Copilot Enterprise. To stop all spend beyond the included pool, an administrator turns it off in the enterprise's or organisation's AI Controls settings. Usage is then blocked when the pool runs out, whatever the budgets say.
• Enterprise policies. Enterprise owners open AI controls at the top of the enterprise page, then Copilot, to set billing, usage and model policies for every organisation, or delegate a decision to organisation owners.
• Organisation policies. Organisation owners open Settings, then Copilot, then Models, to choose which models beyond the basic ones are available. GitHub notes those models may incur additional costs. An enterprise setting cannot be overridden by an organisation.
• One licence per person. A user licensed by several organisations is billed to one chosen at random each cycle, so their spend can land on a different organisation's budget from month to month.

TakeawayUse the paid usage policy as the on or off switch for metered spend, and model policies to keep expensive models for the people who need them.

07How to limit Claude Code costs alongside Cursor and Copilot.

Claude Code takes the limits of the account each developer signs in with: workspace limits in the Claude Console, member limits in claude.ai admin settings on Team and Enterprise, or your cloud's budget controls on Amazon Bedrock, Google Cloud or Microsoft Foundry. Our Anthropic guide covers each route step by step.

TakeawayList how every developer signs in to Claude Code, because each sign-in method has its own limit screen.

08How to set a per-developer monthly budget for AI coding tools.

AI coding tools cost per developer is a seat price plus metered usage, and the usage part varies widely from one person to the next. GitHub's included allowance is worth 19 US dollars a month for each Copilot Business licence and 39 for each Copilot Enterprise licence. Anthropic puts the average Claude Code cost across enterprise deployments at around 150 to 250 US dollars per developer per month.

That spread is why a single company-wide number fails. Set a default per developer in each tool, then give the people who need more a written exception.

1. Set the default. A Copilot universal user-level budget a little above the included allowance, a Cursor team limit sized for the seat count, and the equivalent Claude Code limit for each sign-in method.
2. Watch one full cycle. Read Copilot's AI usage dashboard and Cursor's Spending tab to see who reaches the default and why.
3. Name an approver. One person per team decides on exceptions, with a reason and an end date recorded for each one.
4. Grant exceptions that expire. In Copilot, give the developer an individual user-level budget and set it to expire at the end of the billing cycle or on a date. In Cursor Enterprise, add a member override and note when to remove it, since overrides stay until cleared.
5. Use the request queues. Blocked Copilot members can request more, and owners approve under Requests from members by setting a new amount and clicking Approve and increase.

Cursor and Copilot only see their own spend, so the per-developer number above them has to live somewhere else. The essay on corporate AI spending limits covers how Uber and Tesla set a cap for each employee with approval above it.

TakeawayGive every developer a default in each tool, and route anything above it to a named approver with an end date.

09Common mistakes.

• Leaving enforcement off in Copilot. Stop usage when budget limit is reached is off by default on enterprise, organisation and cost centre budgets, so they only send alerts.
• Leaving Cursor's admin-only toggle off. Any team member can then raise the team limit or turn on-demand usage back on.
• Treating alerts as limits. Cursor spend alerts and Copilot threshold alerts send emails and leave usage running.
• Raising a Cursor limit late in the cycle. On legacy plans that can turn credited overage into a charge.
• Expecting the Copilot enterprise budget to cover licences. It caps metered charges only.
• Forgetting overrides. Cursor member overrides and Copilot individual budgets without an expiry stay in place after the reason for them has gone.

These limits sit inside each vendor's account. Seeing the total across coding tools and model providers is the job of AI spend management, which our guide covers.

TakeawayTurn on enforcement everywhere, and give every override an end date.