Every AI call on one balance.
Capped before it runs.
Every AI action on your Xuda account draws from one shared credit balance: dashboard chats, agents, field assist, generated images, transcription and realtime voice. Put a ceiling on the account, on a workspace profile, on a teammate or on a single model, and Xuda checks it before the model runs.
Counted over the current billing period. It resets on 2 September.
Sends an alert. Nothing is blocked.
Every AI call on the account stops here.
AI models on one balance: text, image and voice
Ways to cap spend, from the account down to one model
Overage to reconcile: the check runs before the model
Days of usage history, down to the individual call
The bill is not the problem. Finding out too late is.
AI usually gets billed after the fact, one vendor bill for the whole team, with no way to tell which person, which project or which model spent it. Xuda meters every AI call as it happens, puts it all on one balance, and checks your limits before the model runs. You do not reconcile an overage. There is no overage to reconcile.
Metered per call. Checked before the call.
Nothing here runs after the fact. Every AI action is priced from the model catalog as it happens, the balance is recomputed from one function, and the gate answers before any model work starts.
- 01
Every call is metered
Xuda writes one usage record per AI call: the model, the action, the tokens in and out, the workspace profile and the person who ran it. Nothing is estimated at the end of the month.
- 02
One balance, one formula
Your plan allowance for the current billing period, plus any credits added on top, minus what the period has spent. The meter, the alerts and the limits all read that same number.
- 03
The gate answers first
Before a model runs, Xuda checks freeze, then the model allow or deny list, then the caps on the account, the profile, the person and the model. A hard limit refuses the call. A soft limit lets it through and tells you.
- 04
You watch it land
The balance in the dashboard updates as calls complete, and the history keeps the individual calls, so any figure on the page can be taken apart.
Six controls, one page, one save.
An account ceiling, a budget per workspace profile, a budget per teammate, a cap per model, allow and deny lists, and a freeze switch. When a profile limit and a member limit both apply, the tighter one wins.
An account-wide ceiling
Set Warn me at and Stop at on the account pool, in credits or as a percent of the allowance. Stop at refuses every AI call on the account once it is reached.
A budget per workspace profile
Give each shared profile its own warn and stop limits, so one busy project cannot quietly eat the whole month.
A budget per teammate
The same two fields, per person. Member limits sit on top of whatever their profile allows, and the tighter of the two is the one that applies.
Model allow and deny lists
Choose allow-list or deny-list on a profile or a person and tick models from the catalog. A call using a blocked model is refused with the reason.
A cap on one expensive model
Set warn and stop credits against a single model, with that model's spend shown beside it. The cheap models stay open while the costly one stays bounded.
An instant freeze
One switch on a profile or a person. Every AI call from there is blocked until you turn it off, and the card says exactly that.
Alerts before the wall
Pick 50, 75, 80, 90 or 100 percent, delivered in the app, by email, or both. A soft breach reports and never blocks.
Where the credits went
Spend by model, by profile and by member, a day-by-day chart, and the recent calls that produced the total.
Top up without changing plan
When a month runs hot, credits can be added on top of the allowance instead of moving plan. They land in a pool the billing period does not clear.
Give the budget an owner.
Open the Profiles tab and set a limit on a shared workspace profile. Open the Members tab and set one on a person. Limits follow the person, on top of whatever their profile allows, and the tighter of the two is the one that applies.
- Profiles and Members are two tabs of the same editor. Every card carries its own warn, stop and freeze.
- Member limits stack on top of profile limits, and the tighter of the two is the one that applies.
- Rules saved against the main profile are discarded on purpose, because the main profile is not shareable.
- A teammate who is not the account owner sees the same page marked View only, with every control disabled.
A number you can take apart.
Press View history on the pool, on a profile or on a person. The dialog opens on the last 30 days and switches to 7 or 90. Xuda writes one record per AI call, with the model, the action, the tokens in and out, the profile and the person, so the total on the card is always the sum of calls you can read.
- Day by day spend, plus credits used, AI calls and busiest day.
- What the credits went on, broken down per model.
- Recent calls with the action, the person and what it cost.
| GPT-5.6 Terra | dashboard chat | Dana K. | 12:04 | 6.2 |
| GPT Image Mini | create image | Amir S. | 11:51 | 4.0 |
| GPT-5 Nano | field assist | Amir S. | 11:47 | 0.3 |
| Whisper Speech AI | transcribe | Dana K. | 10:22 | 1.1 |
| GPT-5.6 Luna | release notes | Noa B. | 09:58 | 2.4 |
Credits you bought are not credits you rented.
Your plan allowance belongs to a billing period. Credits you add on top do not. Xuda keeps them apart in the ledger and still shows you one balance.
The plan allowance resets
Each paid invoice mints the allowance for that billing period, stamped with that period's own start and end. It is counted over the period and it does not roll over.
Credits you add do not
One-time packs, referral bonuses and credits added by support go into a separate pool that the period roll never touches. They sit there until they are spent.
Unlimited stays unlimited
On an unlimited plan there is nothing to gate. Percentage limits resolve to no cap at all and the account is never blocked on balance.
A warning is not a wall.
Warn me at sends an alert and lets the call through. Stop at refuses the call, with a reason that names the scope that stopped it, so nobody has to guess whether it was the account, their profile, their own limit or the model they picked. Pick your thresholds at 50, 75, 80, 90 or 100 percent, in the app, by email, or both.
- Limits in credits, or as a percent of the allowance so an upgrade does not invalidate them.
- Freeze is instant and explicit: every AI call from that profile or person is refused until you turn it off.
- The balance in your browser updates as calls land, without a refresh.
Every plan comes with credits. Team is where you get to cap them.
Your membership plan decides how many AI credits land each billing period. From Team upwards you also get the manager: the account ceiling, the budgets per profile and per person, the model rules and the freeze switch. One-time packs work on any plan, and the AI Workspace add-on stacks more credits onto any paid one.
Free
Start hereRun real AI actions, and watch what each one costs you.
- 30 starter credits on signup, in the pool a billing period never clears.
- A credit every 24 hours rather than one monthly grant, because the free allowance is a drip.
- One credit for every contact connection somebody accepts.
- The live balance, the usage charts and the full history drill-down.
- Credits a billing period
- 30
- AI Workspace add-on credits
- Credit management editor
- Usage history
- Up to 180 days
- One-time credit packs
Pro
A monthly allowance, and add-on credits that stack on top of it.
- An allowance minted the moment the period's invoice is paid, stamped with that period's own window.
- The AI Workspace add-on stacks another 200 to 10,000 credits a month on top of the plan.
- AI image generation in the website builder, once a workspace credit plan is on the account.
- Credits a billing period
- 600
- AI Workspace add-on credits
- +200 to +10,000
- Credit management editor
- Usage history
- Up to 180 days
- One-time credit packs
Team
Where the limits unlockThe whole credit manager: ceilings, budgets, model rules and freeze.
- An account-wide Warn me at and Stop at, in credits or as a percent of the allowance.
- A budget on every workspace profile and every teammate, and the tighter of the two wins.
- Model allow and deny lists, plus warn and stop credits against a single model.
- Freeze a profile or a person on the spot, refused before the model runs.
- Alerts at 50, 75, 80, 90 or 100 percent, in the app, by email, or both.
- Credits a billing period
- 1,600
- AI Workspace add-on credits
- +200 to +10,000
- Credit management editor
- Usage history
- Up to 180 days
- One-time credit packs
Agency and Enterprise
Bigger allowances, up to an account that is never gated on balance.
- Enterprise runs 10,000 then 50,000 credits a period on the first two tiers.
- The top tier is unlimited: no ledger grant, nothing to count, and percentage limits resolve to no cap at all.
- Support can grant up to 100,000 credits to an account in a single go, into the pool that never resets.
- Credits a billing period
- 3,500 to unlimited
- AI Workspace add-on credits
- +200 to +10,000
- Credit management editor
- Usage history
- Up to 180 days
- One-time credit packs
Credit management is a Team plan feature and above. Below Team the page explains what it does and points at the plans, while reading your own balance, usage and history is not gated on any plan. An allowance is minted when the billing period's invoice is paid and is counted over that period, so it does not roll over, while credits added on top sit in a separate pool that never resets. The figures above are credit allowances, not prices: plan pricing is confirmed on the plans page in the dashboard. The redesigned credit manager and the one-time packs are running on our development environment and are still being rolled into the deployed dashboard.
Most teams find out what AI cost them at the end of the month.
These are the three things people actually do instead: read the vendor's usage page, buy a seat per person, or track it by hand in a sheet. Here is what each one gives you.
Xuda AI Credits You | Your AI vendor bill | Per-seat AI tools | Track it in a sheet | |
|---|---|---|---|---|
| One balance for text, image and voice | Per API key | Per seat | Whatever you log | |
| Checked before the model runs | ||||
| Account-wide ceiling you set yourself | Budget alerts | Seat count | ||
| Budget per workspace profile | By hand | |||
| Budget per teammate | Add or drop a seat | By hand | ||
| Cap on one expensive model | ||||
| Allow or deny specific models | ||||
| Freeze a profile or a person on the spot | Remove the seat | |||
| Alerts at 50 to 100 percent, in app or email | Threshold emails | |||
| Spend traced to the individual call | Per key, per day | Whatever you log | ||
| Top-up credits | Never expire | Varies by vendor | Not applicable | Not applicable |
| Best for | Teams sharing one AI budget | One developer, one key | A fixed cost per person | Very small teams |
Answers before you have to ask.
Put a ceiling on AI spend before you need one.
One balance for every AI call on the account, limits checked before the model runs, alerts before the wall, and a history that traces every credit back to the call that spent it.