Spend cap and credits

A War Room council is metered in GNG credits, the same in-product unit as the rest of the AI Console. You set a per-room spend cap before you start, a live readout tracks the running spend against it, and the finished council reports its total, its per-seat spend, and its cost in your history. This page explains the spend cap control, what raises a council's cost, where every cost figure appears, and what happens when the cap is reached. The War Room is available on the Pro plan and above.

Before you begin

Use a Pro plan or above. The War Room is rolling out to eligible members in waves, so an eligible account may not see it yet.
Understand the shared credit meter first. A council draws down the same balance as the rest of the AI Console. For what a credit is and how the meter refills, see GNG credits.
Costs in the War Room are always shown in GNG credits, never in currency or token counts.

Quickstart

Set your topic and seat the table
Write a topic and seat at least two panelists so the room is ready to start.
Open the settings panel
In step 3, Format, open War Room settings, where the spend cap lives.
Set the spend cap
Drag the Spend cap slider or tap a preset amount, from 25 up to 20,000 credits. The default is 250.
Start the session
Select Start session and watch the live Round, Spend, and Cap readout as the council works.
Read the totals
When the council finishes, review the Total, the number of Rounds and Seats, and each seat's own spend.

What it shows

The cost of a council appears in several places, from the control you set before you start to the record it leaves in your history. The surfaces are described in the following list.

Spend cap control (in Format): a hard ceiling in GNG credits, set with a slider and a row of one-tap preset amounts. Its caption reads "Hard ceiling on this room's spend. Hitting it ends the room and delivers a closing summary."
Live spend readout: while the council runs, a status panel shows the current Round, the running Spend, and the Cap, all in GNG credits. The Spend figure turns amber as it nears the ceiling.
Header balance: a compact readout in the header shows the running session spend and your remaining credit balance.
Final totals: when the council finishes, the panel shows the Total spend, the number of Rounds, and the number of Seats, and it lists each seat with the credits it spent.
History cost: the Cost column in Past councils records the total credits each session used.
Design cost line: when you use Design panel with AI, it reports the credits it spent when it finishes, so the setup step is accounted for before the debate begins.

Inputs

The only cost control you set directly is the spend cap. The other settings that change a council's cost live in the table and in Format, and are covered on their own pages.

Spend cap: a hard ceiling in GNG credits. The slider ranges from 25 to 20,000, with a default of 250. One-tap presets are 250, 500, 1,000, 2,500, 5,000, 10,000, and 20,000. The slider snaps to clean steps: 25 under 500, then 100 up to 1,999, then 500 up to 20,000. Above roughly 60 percent of the range a caution reads "Big ceiling. The room will keep arguing until it lands here."

What raises a council's cost, set elsewhere before you start, is described in the following list.

More seats: each seated panelist speaks every round, so a larger table costs more per round.
More rounds: each added round is another turn for every seat. Optimal rounds let the host stop when it has heard enough.
Simultaneous round style on a large table: firing every seat at once each round can raise cost faster than rotating one speaker at a time.
Optimal model mode: the toggle "Let the orchestrator promote each agent to its best-fit model for this topic" costs more credits per turn. Off by default.
Enable all tools for subagents: giving every seat the full research toolkit can add cost through more tool use. Off by default.
Heavier per-seat models: a more capable model chosen from the picker costs more per turn than a lighter one. Leaving a seat's model unset lets the room pick a cost-efficient default.

Outputs and how to read them

Read the running Spend against the Cap to see how close the council is to stopping. The amber tint as the spend nears the cap is your near-the-ceiling cue.
When a council finishes normally, the Total is the sum of every seat's turns plus the host opening each round and writing the closing synthesis. The per-seat list shows where the credits went.
When a council reaches its cap, it stops gracefully. The status reads Session limit reached, shown in amber rather than as an error, and the host still delivers a closing summary. The stop message itself does not print a number.
A row's Cost in Past councils is the total credits that session used, so you can compare setups at a glance.
Sharing a council, exporting its transcript, and replaying a finished session cost nothing. Only starting or continuing a session draws down credits.

How it is calculated

There is no separate scoring formula behind a council's cost. In plain terms, a session's cost is the sum of its parts: each seat's turns across the rounds, plus the host opening each round and writing the synthesis. The platform meters each part and shows the running and final totals directly, always in GNG credits and never as currency or token counts. The settings that raise the total are more rounds, more seats, Simultaneous style on a large table, Optimal model mode, enabling all tools, and heavier per-seat models. Because the cost depends on those settings and on how much the panelists write and look up, the same topic can cost different amounts on different setups. For the shared credit mechanic and the per-answer breakdown, see GNG credits.

Data and timing

The live spend updates as the council runs, and the final totals settle when the session ends. The header balance reflects your live credit balance. A per-room spend cap applies only to that one council; it does not change how your account balance or refills work. The cost mechanic itself involves no market data feed; the live company and market data the panelists pull while they debate is a separate step, current to the platform's normal refresh schedule.

Use cases

Set a modest cap for a quick two-seat check, and a larger cap for a deep multi-round debate.
Watch the running spend and end a council early once you have the answer you need.
Check the per-seat breakdown to see which panelist drove the cost.
Lower the rounds, the seats, or the per-seat model to make a credit balance last longer.

Limitations and disclosures

The exact cost of a council depends on its settings and on how much the panelists write and look up, so no fixed figure applies to a given topic. Availability is by plan: the War Room is on Pro and above and is rolling out in waves, so an eligible account may not see it yet. The cost mechanic itself carries no investment-advice statement; where a spend view sits beside a verdict, that verdict is research support, not a recommendation.