Billing and usage
What Mobius costs, where the credits went, and how to stop a job spending more than you meant.
Mobius bills your organization for the AI usage and platform work it does. In a direct-billed organization, open Settings > Billing to see the plan, remaining credits, and invoices, and Settings > Usage to see the shared pool, per-person limits, and itemized charges. A managed organization has only Settings > Usage, because its service provider owns the payment relationship.
The organization owner can change plans, buy credits, open the billing portal, and edit usage limits. Other roles with billing read access see the figures without those controls; an organization Admin does not receive billing mutation authority merely from the Admin role.
Reading the pages
Three places:
| Where | Answers |
|---|---|
| Billing > Overview | What plan am I on, how many credits are left, is anything wrong? |
| Billing > Plans | What plan should we have, and what have we paid? |
| Usage | Who is using Mobius, what's spending the credits, and what limits apply? |
Available credits is what Mobius can spend right now, and it comes from two places:
- Included with your plan, which resets each billing period and is spent first.
- Your credit wallet, which holds credits you bought or were granted. It's spent after the included ones and carries over until used.
The billing period is the calendar month, from midnight UTC on the 1st to midnight UTC on the next 1st. An organization that starts mid-month gets a short first period that runs to the next 1st: it's charged for the days it actually has, and its included credits are sized for those days, rounded up to whole weeks. Every period after that is a whole month.
The top navigation and the Overview card show the total. The Credits card breaks it into both sources. This trips people up: a wallet at zero doesn't mean you're out of credits, if this period's included credits are still there.
Where did the credits go?
Open Usage and start with Recent credit charges. Each row names what was used, which AI model where relevant, how much, and what it cost. Follow the row through to the run behind it.
Credit movements is the wallet ledger. Purchases and grants add to it. Metered work draws down your included credits first, so the ledger only records the part that spilled into the wallet. Work that cost nothing still shows up in the usage table, so you can tell "we did this and it was free" apart from "we didn't do this".
If you bill clients for this work, one organization per client is what makes the usage breakdown line up with an invoice.
Put limits at the right boundary
Open Settings > Usage to see the monthly shared pool, what each person has used, and each person's effective credit limit. The organization owner can set a Default person limit for the current billing period and override it for a specific member. A blank limit means no per-person ceiling; it does not make the shared pool unlimited.
For a direct-billed organization, the owner can also choose what platform-funded work does when the shared pool and wallet are empty: stop, or use the configured economy model. In a managed organization, the service provider controls that pool-exhaustion policy while the organization owner can still tighten the local default person limit.
Those period controls answer who may consume the organization's credits. Put the narrower boundary on the work too: each routine has Credits per run and Credits per 24 hours, and each assistant turn has an active-work timeout. Review those guardrails before scheduled or event-triggered work goes live.
After a payment or plan change
Mobius may show the operation as processing for a moment while the payment provider confirms. Leave the billing page open; it checks and refreshes the subscription, credits, usage, and invoices once things settle.
If you come back from checkout without paying, Mobius cancels the pending operation where it can. If your payment went through at the same moment, it finishes reconciling rather than falsely telling you it was cancelled.
Next
- Cap spend with guardrails.
- Trace charges back to the conversation in agent sessions.
- Limit what agents can reach with tools and skills.
- Split usage by client with separate organizations.