Control Center

The Control Center (admin sidebar → Billing & Control → Control Center) shows who can use Ronja’s AI, how much they can spend, and what happens when a limit is reached. It has three tabs:

People: every member, what they can do right now, and why.
Rules: what each limit does, and every rule in one table.
Spend response: what happens as the organization uses its credits, and the default models. Needs billing access.

Every change is made in a dialog and saves when you confirm it. What credits are is covered in Credits and AI spend.

What each limit does

Ronja has four separate limits. Each counts different spend and does different things as it fills up. The Rules tab opens with the current week’s dates (and, with billing access, the billing period’s), as on People, then this table, worked out from your organization’s own settings. Without billing access the table has two columns, the daily AI safety limit and the weekly budget, and opens with “Two separate limits.” Above it, one line says which spend the organization’s steps measure, with your own steps’ percentages, for example: “The 70%, 85% and 95% steps measure the whole organization’s credits for the billing period (monthly). They never apply to a person’s weekly budget.”

	Organization credits	Daily AI safety limit	Weekly budget	Billing-period allocation
Counts	All credit use in the organization, automations included	All AI use in the organization, automations’ agents included, counted as on your bill	One person’s chats with Ronja	What one person runs: chats, Slack/Teams, API and CLI calls, older automations
Applies to	Shared by the whole organization	The whole organization	Each person, admins included	Each person except admins. Blocked applies to everyone.
Before the limit	The steps you set on Spend response, for example an email at 70%, a warning to members at 85% and reduced models at 95%	At 80% admins get a heads-up and members see a note in the chat	At 75% the person is warned in the chat	No warning
Switches models	Only at a reduced-models step, and only where reduced models are set	Never	Never	Never
At 100%	What running out of credits does (see Spend response)	AI pauses for everyone until the reset	That person’s chats stop	That person’s chats stop
Resets	Every billing period (monthly), 00:00 UTC on your billing date	Every day, 00:00 UTC	Every Monday 00:00 UTC	Every billing period (monthly), 00:00 UTC on your billing date

The daily AI safety limit is the organization’s safeguard against an unexpected spike in AI use, such as a runaway agent or loop: when it is reached, AI pauses for everyone until 00:00 UTC. It is not a billing problem, and it counts credits as on your bill, so its numbers aren’t directly comparable with a weekly budget’s. See Daily AI safety limit and, below, Daily AI safety limit on Rules.

A weekly budget and an allocation are per person: each person gets their own amount, and one person using theirs doesn’t reduce anyone else’s. Weeks run from Monday 00:00 UTC to the next Monday 00:00 UTC. The billing period runs from your billing date to the same date the next month. The top of the People and Rules tabs shows the current windows once: This week “5–11 Oct · resets in 4 days” and, with billing access, Billing period “1–31 Oct · resets in 24 days”. People also has a Today line for the daily AI safety limit, for example “62% of the daily AI safety limit · resets at 02:00”, shown in your own time. East of UTC the reset falls in the small hours, so for most of the day it reads “resets tomorrow at 02:00”. Spend response shows the billing period the same way. The dates are the UTC days each window covers, last day included, and the days until the reset are counted on your own calendar. If your clock is behind UTC, as in the Americas, the dates read “5–11 Oct (UTC)”: for you, a new UTC day starts the evening before.

If your organization is on a monthly budget, the table has a Monthly budget column in place of Organization credits (without billing access, it comes first and the table opens with “Three separate limits.”). It counts “All usage in the organization, automations included, counted as on your invoice; the platform fee is not included”, is shared by the whole organization, emails admins at 80% (80% email to admins, followed by your spend steps), and at 100% “New chats and runs stop for everyone, admins included, until the next billing period or until an admin raises it”. It resets every billing period. If Ronja has set no budget, the column says “No email or step runs without a monthly budget” and, at 100%, “No monthly budget: nothing stops.”

On a phone, the table becomes one block per limit.

Weekly budgets and allocations stop chats with Ronja only. Automations, Saved Agents and Slack or Microsoft Teams conversations are not stopped by them. The daily AI safety limit pauses all AI, automations’ agents included.

When more than one limit stops someone at once, the organization’s credits running out (or its monthly budget being used up) comes first, then the daily AI safety limit, then the person’s own weekly budget or allocation.

Which rule applies

For each setting the most specific rule wins: Personal exception › Access group or role › Default for each person.

Weekly budget: a personal exception wins, even when it is lower. Otherwise a person gets the highest budget among their access groups, and No limit beats any number. A group budget applies even when it is lower than the default, so Interns can have 25 credits while everyone else has 100. Weekly budgets have no role rule. See Set weekly AI budgets.
Billing-period allocation: a personal exception, then the person’s role, then the default. With nothing set there is no limit. A Limited allocation never stops an Admin; Blocked stops everyone.
Build mode and Choose model & effort: a personal exception, then the role, then the default, which is always On for build mode and Allowed for model choice. Two roles are fixed and win over any exception: Super Admins always have build mode and model choice, and User Read-Only members never do.

People

The People tab lists every member of the organization, with each person’s email under their name so two people with the same name can be told apart. Search with Search people, which matches a name or an email, or filter with All, Stopped, Near a limit and Has an exception. While AI is paused by the daily AI safety limit, a Paused for today filter appears too. The columns are Person, Status, Weekly budget, Billing period, Build mode and Picks model; without billing access you see Person, Status and Weekly budget only. Under each budget, a small line says which rule it came from: Personal, an access group’s name, a role such as Role: User, or Default (the default for each person). Above the list, Today says how much of the daily AI safety limit the organization has used today, This week gives this week’s dates and when it resets, and with billing access Billing period does the same for the billing period. On a phone the list becomes one card per person.

For an organization on a monthly budget, a line under the dates says how much of it is used, for example “This billing period: 62% of the monthly budget · resets 3 Nov”. Once it is used up, a grey strip at the top of the tab says “Your organization has used its monthly budget. New chats and runs stop for everyone and resume when the next billing period starts on 3 Nov, or as soon as an admin raises the budget under Rules.”

While the daily AI safety limit has paused AI, a grey strip at the top of the tab says so, for example “AI is paused for everyone until tomorrow at 02:00 — daily AI safety limit.”, and who can change the limit: “Your organization’s own limit applies. You can change it under Rules.” or “Ronja’s limit for your organization applies. Contact Ronja support if you need it raised.”

The Status column says what happens when that person sends a prompt now:

Can chat; without billing access, Within weekly budget or No weekly limit, because only the weekly budget was checked.
Near weekly limit or Near period limit: they have used 75% or more. An Admin is never near an allocation, because a limited allocation doesn’t stop them.
Weekly budget used up, Allocation used up or Chats blocked: their chats are refused.
AI off for Read-Only: they are User Read-Only and your organization has turned AI off for read-only members.
No role: they can’t use Ronja in the app until someone gives them a role in Users & Access.
Credits ran out: the organization has used all its credits, so everyone’s chats are stopped.
Monthly budget used up: the organization has used its monthly budget, so everyone’s chats are stopped until the next billing period or until an admin raises it.
Paused for today: the organization reached its daily AI safety limit, so AI is paused for everyone until 00:00 UTC. It resumes on its own, and it isn’t a billing problem.
Couldn’t check: a setting their status depends on couldn’t be read. Try again in a moment.

A person’s panel

Click a person to open their panel beside the list (below it, on a phone). Its link opens the same panel, so you can share it. The panel starts with one sentence about what they can do, then has a section for each setting: Daily AI safety limit, Weekly budget, Billing-period allocation, Build mode and Choose model & effort. The Daily AI safety limit section is the organization’s, the same for everyone: it shows how much of today’s limit is used (or Paused for today), and its Daily AI safety limit in Rules link opens the setting.

Each section lists the rules from most specific to least, ending with Default for each person. When no weekly budget or allocation is set anywhere, a last Nothing set rule applies: No limit. For build mode and model choice, a fixed role comes first: Locked role · Super Admin reads Always on for build mode and Always for model choice, and Locked role · User Read-Only reads Always off and Never. The one that applies is marked, and the others are struck through. A section that couldn’t be read shows a single Couldn’t read line instead, with nothing marked. Under a limit, the panel says what happens as it fills up and names the day it resets, for example: “At 75% they are warned in the chat. At 100% their chats stop until the week resets on Mon 12 Oct. This limit never switches their model.”

With billing access, the Organization section shows how much of the organization’s credits are used this billing period, as a percentage, and what the spend steps will do. These steps use the whole organization’s credits, not the person’s own budgets.

Change a person’s rules from their panel:

Click Add exception under a setting (or Edit exception or Remove exception when they already have one).
Choose the value. A weekly exception takes Credits per week or No limit. An allocation exception takes Credits per billing period, No limit or Blocked. Build mode takes On or Off, and model choice Allowed or Not allowed.
Enter a Reason. It is required for weekly and allocation exceptions, and is kept with the exception.
Click Add exception (or Save exception).

When a person’s role decides build mode or model choice on its own, the panel says so: a personal exception would not change it. If their exceptions couldn’t be loaded, the build mode and model choice buttons wait and the panel says “Couldn’t load this person’s exceptions. Try again.”, so a save never overwrites the setting it couldn’t see.

Give extra credits

When someone runs out of their weekly budget before Monday, you can top them up for the rest of the week:

In their panel, click Give extra credits this week. It appears for anyone whose weekly budget has a limit.
The dialog shows how much they have used this week. Extra credits starts at 50; change it if you like, up to 1,000,000 at a time.
Enter a Reason. It is recorded in the audit log.
Click Give credits.

Extra credits count until Monday 00:00 UTC and then disappear. To take them back sooner, click Remove extra credits in the person’s panel, or the same button next to their name under This week only on the Rules tab. It removes all of this week’s extra credits for that person at once. They only lift the weekly budget, so they won’t help if the person’s billing-period allocation or the organization’s credits are what stop them. When someone uses up their weekly budget, your admins get an alert that links straight to this dialog for that person. See Set weekly AI budgets.

Rules

Under What each limit does, the Monthly budget section (where your organization has one) and the Daily AI safety limit section, the Rules tab shows every rule in one table: a row per setting (Weekly budget, Billing-period allocation, Build mode, Choose model & effort) and a column for each layer (Default for each person, By role or access group, Personal exceptions). In the first column each rule is named Default. Without billing access, only the Weekly budget row is shown.

Click a value, such as 100 / week or Default (On), to change it. A value you can change is underlined; a value with a lock can’t be changed.
The Weekly budget row says “Each person gets their own amount every week. It isn’t a shared pool: one person using theirs doesn’t reduce anyone else’s.” The current dates are at the top of the tab, not in each row.
Set a group budget gives an access group its own weekly budget. When a person is in several groups with a budget, the row says which one applies to them.
This week only lists the people who have extra credits this week. Remove extra credits takes them back.
Add exception gives one person their own rule; the person list shows each person’s email beside their name. Remove deletes it.
A role with no rule of its own reads Default (for build mode and model choice, Default (On) or Default (Allowed)); in the role’s dialog, Use the default for each person (or Use the default (On)) removes its rule.
A rule on a role that is no longer active, or no longer exists, stays listed with (inactive) after the role’s name, so you can still change it or remove it. Members who still hold an inactive role keep getting its rule.
In the role column for build mode and model choice, Super Admin always reads Always on (or Always) and User Read-Only Always off (or Never). These can’t be changed, and neither can the default for each person (always On for build mode, Allowed for model choice).

Build mode lets a person create and edit workflows, apps and notes. Deep-analysis chats don’t need it; see Standard, Deep analysis, and Build. Choose model & effort lets a person pick a model and effort per chat instead of the organization’s default. It is a permission, not a spending limit. While the organization is out of credits, nobody can pick a model; on a monthly budget, the line reads “While the monthly budget is used up, nobody can pick a model.”

Monthly budget

If your organization is on a monthly budget, a Monthly budget section follows What each limit does: “When it is used up, new chats and runs stop for everyone until the next billing period. Counted as on your invoice; the platform fee is not included.” Every Admin sees it, with or without billing access. Its row, Your organization’s budget for each billing period, shows the budget, for example “1,000 credits”, and “This billing period: 62% used · resets 3 Nov”. To change it:

Click Edit. If Ronja has set no budget, the row reads “No monthly budget. Nothing stops when usage grows.” and the button is Set budget.
Enter Credits per billing period, a whole number of credits up to 1,000,000.
Click Save.

The change applies within about 20 seconds, so raising the budget lifts the stop straight away. There is no Remove: only Ronja can take the budget away. It can be changed only in the browser, never by an API token, the CLI or an automation.

Daily AI safety limit

Between What each limit does and the rules table, Daily AI safety limit holds the organization-wide limit. When today’s AI use reaches the lower of its two limits, AI pauses for everyone in the organization until 00:00 UTC. Every Admin sees it, with or without billing access. It has three rows:

Your organization’s daily AI safety limit: optional. Not set, or the number of credits per day, with Set a limit, Edit and Remove.
Ronja’s limit for your organization: set by Ronja, read-only. Contact Ronja support to have it raised.
Today: how much is used, for example “62% used · resets at 02:00”, or “Reached: AI is paused for everyone until 02:00”.

In effect marks the lower limit, the one that applies. To set your own:

Click Set a limit (or Edit).
Enter Credits per day, counted as on your bill, at or below Ronja’s limit.
Click Save limit.

Remove, then Remove limit, leaves Ronja’s limit alone in effect. A change applies within a minute, so raising the limit lifts a pause without waiting for the reset. It can be changed only in the browser, never by an API token, the CLI or an automation. If Pausing is switched off by Ronja right now shows under Today, reaching the limit pauses nobody for the moment. See Daily AI safety limit for what everyone sees during a pause and the notices admins get.

Spend response

Spend response opens with This billing period, the organization’s credits for the current period. One line gives the period’s dates and says the credits are shared, for example “1–31 Oct · resets in 24 days. Shared by the whole organization.” Below it are how many credits are used and a meter marked with your steps. Usage report opens the Usage page.

What happens as the organization spends lists the steps. Its header says which spend they measure, with your steps’ percentages, for example “The 70%, 85% and 95% steps measure the whole organization’s credits for the billing period (monthly). They never apply to a person’s weekly budget.” People’s allocations never trigger them either, and neither a weekly budget nor an allocation switches models. A step is one of:

Email a warning: emails the people chosen in Notification routing. Until you choose recipients there, the step reads “Nobody receives it yet.” with a link to fix it.
Warn members in chat: every member sees a low-credits warning above their chat.
Switch chats to the reduced models: new chat turns use the reduced models, and a setting without a reduced value keeps its standard one. A member’s own model or effort pick is kept. Automations, Saved Agents and Slack/Teams are not affected. Only one of these steps can be on at a time.

Click Add step, choose What happens, enter At this % of the organization’s credits (1 to 100) and turn Step is on on or off. Edit and Remove change an existing step. The steps only run when your organization’s credits are enforced and funded, or it has a monthly budget, whose percentages they then measure; otherwise the tab says so. A step that has fired this period reads Reached this period.

The last row, Credits run out, says what running out does in your organization. Usually chats that would use Opus or GPT-5.6 Sol first use Sonnet instead, and nobody can pick a model; past the overdraft agreed with Ronja, chats stop for everyone, admins included. Automations and Saved Agents stop there too, or at a deeper limit if Ronja has set one. On a monthly budget the row is Monthly budget used up instead: “When the monthly budget is used up, new chats and runs stop for everyone, admins included, until the next billing period or until an admin raises it. Work already running finishes.” If Ronja has set no budget, it reads “No monthly budget: nothing stops.”

Default models

Default models sets what chats use unless a member picks their own model, for two kinds of chat:

Analyst: questions and exploration (Standard chats).
Build & deep analysis: Build mode and Deep analysis chats.

Click Edit on a row to set its Standard model, Standard effort, Reduced model and Reduced effort. A reduced value left as Same as standard keeps the standard one during the reduced-models step. The model list marks a less proven model as “experimental”. Claude models show the version running now, for example Opus 5.5; your choice follows new versions of that model.