Agent Studio
Quota Control
Path: Settings → Quota Settings
Quota Control throttles how many tokens a user can spend, across scopes such as a 60-second burst, a rolling day, a rolling month, and any single conversation. Above all of them sits one ceiling for the agent itself, which counts every user’s usage together.
Each scope catches what the others miss. The burst limit stops automated abuse in the minute it starts. The daily limit stops one user exhausting a month’s allowance in an afternoon. The monthly limit is the one you price a tier against. The conversation limit bounds a single runaway thread. The agent ceiling is the backstop under all of them: when it trips, the agent is hidden from everyone until usage falls back below the limit.
Set against each other, these are a pricing lever as much as a safety limit. Throttling tokens per user, and per user tier, lets you offer the agent to everyone while keeping cost predictable, then lift the ceiling for the users, or the plans, that justify it. If you plan to monetize on AI, quotas are where a free tier ends and a paid one begins.
What a user sees on the way to a limit is configured separately: see Warn users before they hit a limit.
Quota reference
Section titled “Quota reference”Every quota is evaluated on each request, and the first one exceeded triggers its block behavior. The Enforced chain in Quota Settings shows the order, from the broadest limit to the narrowest: Agent, Monthly user, Daily user, Conversation, Burst. When more than one quota is exceeded, that order decides which block the user sees.
| Quota | Window | Platform default | Purpose | Block behavior |
|---|---|---|---|---|
| Agent | 24-hour (rolling) | 200,000,000 | Caps total token usage across all users of the agent. | The agent is hidden until usage falls below the limit. |
| Monthly user | 30-day (rolling) | 33,000,000 | Enforces long-term per-user billing and usage tiers. | Input disabled at 100%, showing the resume date and time. |
| Daily user | 24-hour (rolling) | 6,500,000 | Stops a single user from exhausting their monthly limit in one day. | Input disabled at 100%, showing the resume time. |
| Conversation | Per-thread | 4,450,000 | Bounds the tokens a single thread can spend. | Input disabled at 100%, prompting the user to start a new chat. |
| Burst | 60-second | 600,000 | Rate-limits to prevent automated abuse and spam. | Input disabled briefly. “You’re sending messages too quickly.” |
Quota Settings shows the current platform default next to each limit, with a cost estimate for each.
Override the platform defaults
Section titled “Override the platform defaults”Every agent starts with Custom limits off and inherits the platform default for each quota. Those defaults are enforced, so an agent with custom limits off is still capped.
Turn Custom limits on to override any quota. Each row then gets its own On toggle and an editable token value: leave a row off to keep inheriting its platform default, or turn it on and set your own number. The Platform default / cost column keeps showing the inherited value next to a cost estimate, so you can see what a ceiling is worth before you change it.
Cost estimates are priced against the model the agent runs on, named in Cost estimates based on. The same token ceiling costs more on a more expensive model.
Warn users before they hit a limit
Section titled “Warn users before they hit a limit”Warnings are a separate setting, off by default. With Threshold warning off, users get no heads-up: the first thing they see is the block.
Turn it on to warn users in the chat as they approach a limit. It covers the daily, monthly, and conversation quotas:
| Quota | Warning |
|---|---|
| Monthly user | “You’re nearing your monthly usage limit.” |
| Daily user | “You’re nearing your daily usage limit.” |
| Conversation | “You’re nearing the limit for this conversation.” |
The Agent and Burst quotas have no warning state. Threshold warning is available only once Custom limits is on.
How quotas are calculated
Section titled “How quotas are calculated”Quotas count weighted tokens (input + output), not raw messages. A quota tracks the compute cost of each interaction, so a handful of long, context-heavy turns count the same as many short ones.
Rolling windows
Section titled “Rolling windows”Daily and monthly quotas use a rolling window rather than a calendar reset. The system looks back exactly 24 hours or 30 days, so a user’s oldest usage continuously drops out of the window and access returns gradually. This avoids the traffic spikes that follow a shared midnight or first-of-the-month reset, and gives users a smoother experience than a hard cutoff.
Configure quotas
Section titled “Configure quotas”Quotas are configured from Settings → Quota Settings: pick an agent in the selector, then set its limits for the quota types above. Each agent runs on its own budget, so you can give a high-value agent more headroom than a public one, all from your own cost modeling.
Related
Section titled “Related”- Cost & tokens: the events that record what agents actually spend, by action and model.
- High Availability: a separate concern, keeping the agent answering when a provider degrades.