AI Credits & Budgets
How the AI assistant and Community Pulse are billed: what a message costs, where the money comes from on each plan, how to cap assistant and Community Pulse spend separately, per-source Community Pulse budgets, and what happens when credits or a cap run out.
The AI assistant is pay-as-you-go. Each message you send is charged for the AI work it takes, and the charge is paid from your organization's credits. You can cap how much the organization spends on AI each billing cycle.
Credits and caps are per organization, not per user or per project. Everyone in the organization draws from the same credits, and the assistant in every project counts toward the same cap.
MCP connections don't use credits. There, the AI model runs in your own tool (Cursor, Claude, and so on) and is billed by that tool's provider. Gamebeast doesn't charge for the calls it answers.
What a message costs
AI models measure work in tokens, which are small chunks of text. A message is charged for:
- Input: everything the model reads to answer. That's your message, the conversation so far, any context you pinned, and the data it looks up while working. Input the model has read recently in the same conversation is billed at a heavy discount.
- Output: everything the model writes, including its thinking.
The price per token depends on the model. Mini Beast is the cheapest, Beast costs more, and Godfrey costs the most. Some of the assistant's background steps, such as searching your project or building a query, always run on Mini Beast whichever model you picked, which keeps them cheap.
Each message is charged once, after it finishes, and costs at least 1¢. If you stop a reply partway through, you're charged for the work done up to that point.
Keeping costs down
- Start a new conversation when you change topic. The whole conversation is re-read on every message, so a long chat makes each new message cost more. Reloading the page or switching project starts a fresh chat.
- Pick the smallest model that does the job. Mini Beast handles lookups ("what events do I track?"). Save Godfrey for hard analysis.
- Turn thinking down for simple questions. Thinking is billed as output, and output costs the most.
- Pin what you mean with + instead of asking the assistant to search for it. See Attach context.
Where the money comes from
| Plan | How AI is paid |
|---|---|
| Free | A one-time $5 AI credit when the organization is created. It doesn't renew. |
| Beast | Prepaid credits. Add credits whenever you like, or turn on auto-reload. Unused credits carry over between billing cycles. |
| Enterprise | Charged to your credits if you have any. Otherwise it's invoiced monthly with the rest of your on-demand usage. |
Credits are shared by all on-demand usage, so AI and on-demand markers draw from the same credits.
Add credits
Anyone with permission to manage billing can add credits from Billing → Overview:
- Add More tops up once, from $5 up to $50,000, using your saved payment method.
- Auto Reload tops up automatically when your credits fall below a threshold you choose, so the assistant never stops mid-week. It needs a saved payment method. If a reload payment fails, auto-reload turns itself off. Fix the payment method and turn it back on.
If auto-reload is off, we email you when your credits run low (below $10 by default, or your auto-reload threshold if you've set one) and again when they run out.
Enterprise
AI usage never stops because of an empty balance on Enterprise. Usage is charged to your credits, and your credits can go below $0. Any negative amount is invoiced as on-demand usage at the end of the billing cycle and then reset to $0. To put a ceiling on AI spend, set an AI cap.
Set an AI cap
An AI cap limits how much your organization can spend on the assistant in each billing cycle. It covers the assistant only: Community Pulse has its own cap. It's a spending limit, not a separate pot of money. AI is still paid from your credits (or your invoice), but it stops once the cap is reached.
To set one, go to Billing → Overview, then under Caps pick AI Assistant:
- Off (Uncapped), the default. AI spend is limited only by your credits, or on Enterprise, it's invoiced with no dollar limit.
- On, with an amount. AI stops for the rest of the cycle once that amount has been spent.
- On, at $0. The assistant is turned off for the organization.
Things to know about the cap:
- It resets every billing cycle and stays in place until you change it. You don't need to set it again each month.
- Changes apply immediately. Raising or removing the cap makes the assistant available again straight away.
- You need permission to manage billing to change it. Anyone who can view billing can see it.
Alerts
While a cap is set, we email your billing contact and your organization's admins when AI spend reaches 50%, 75%, 90%, and 100% of the cap. You get at most one email per threshold per cycle.
When you run out
The assistant checks your credits and cap before each message:
| Situation | What you see | What to do |
|---|---|---|
| Credits are at $0 or below (Free and Beast) | Your balance is empty | Reload Balance to add credits, or Go to Billing. Free organizations can upgrade. |
| The AI cap for this cycle is used up (any plan) | Your AI budget is exceeded | Manage AI Budget to raise or remove the cap, or wait for the next billing cycle. |
A message that's already running always finishes. If it uses more than your remaining credits, your credits can end up slightly below $0, and the next top-up covers the difference. Until then, the assistant's prompt is disabled.
Community Pulse
Community Pulse uses AI to analyze the messages your sources collect, and that analysis is paid from the same credits as the assistant.
What analysis costs
Community Pulse is charged per message analyzed. Every message a source analyzes is scored by a lightweight model, and messages that carry real player sentiment get a deeper read. Charges are made in small batches as messages are analyzed, not once a month.
Only messages from your production environment are analyzed. Studio and other test traffic is never charged.
Monthly budget per source
Each source has its own monthly budget, which defaults to $50. It's the most that source can spend on analysis in a calendar month (UTC). It resets on the 1st, and it isn't a separate pot of money: spend still comes from your credits.
You set the budget when you create a source and can change it later from the source's Edit panel. Changing it needs permission to manage Community Pulse. Each source shows what it has spent this month against its budget.
When you create a source, we look at the last 7 days of that marker in production and show an estimate: about how many messages a day it receives, roughly what analyzing all of them would cost per month, and how much of it your budget covers. The sample rate slider marks a recommended rate: the share of conversations your budget can cover at that volume. A higher rate doesn't analyze more, because the budget caps analysis at that share.
Community Pulse cap
Community Pulse is billed and capped separately from the assistant. Its spend never counts toward the AI cap, and the AI cap never pauses Community Pulse.
On top of each source's own budget, you can cap what all of your organization's Community Pulse sources spend together in a billing cycle. Go to Billing → Overview, then under Caps pick Community Pulse. It works like the AI cap: off (uncapped) by default, on with an amount, or on at $0 to stop Community Pulse analysis for the organization. Alerts go out at the same 50%, 75%, 90% and 100% thresholds.
The Community Pulse cap covers source analysis, the daily synthesis, and topic previews.
Sampling
If a source receives more messages than its budget can cover, Community Pulse samples it rather than stopping partway through the month. Messages are grouped into short stretches of conversation, and sampling picks whole stretches, so every analyzed message keeps the context around it. Sampling is spread evenly across channels, servers and time, so no single busy conversation decides what Community Pulse sees.
The share of conversation analyzed is recalculated every day from the source's recent volume and its budget, aiming to keep monthly spend a little under the budget. The sample rate you set on the source is a ceiling: Community Pulse never analyzes more than that, even when the budget would allow it. The source shows the share it's currently analyzing.
Including the last 7 days
When you create a source, enroll new Discord channels, or add a topic, Community Pulse offers to include the last 7 days of messages. It's free: it uses no AI credits and doesn't count toward the source's budget or the Community Pulse cap. Nothing is included unless you choose it, and each source, channel and topic can include its last 7 days once. Very busy sources are sampled.
When a source pauses
A source that has stopped analyzing shows why:
| Badge | Why | What happens |
|---|---|---|
| Paused: budget reached | The source has spent its monthly budget. | Analysis resumes on the 1st (UTC), or shortly after you raise the budget. |
| Paused: credits or Pulse cap | Your organization's credits have run out, or its Community Pulse cap is used up. | Add credits or raise the cap from Billing. Raising the source's budget doesn't help here. |
A paused source keeps its place. Messages that arrive while it's paused wait to be analyzed for up to 30 days, so a source that resumes within that window catches up. Messages older than 30 days expire without being analyzed or charged.
This is separate from pausing a source yourself: a source you've paused isn't analyzed or charged until you activate it again.
See what you've spent
Billing → Overview shows:
- Credits: what's left to spend.
- This cycle → AI Assistant and This cycle → Community Pulse: what each has cost the organization so far this billing cycle.
- Caps → AI Assistant and Caps → Community Pulse: each current cap, or Uncapped.
On Enterprise, AI usage appears on your monthly invoice as On-demand AI usage.