How much Codex quota is left?

Where Codex shows it

The usage dashboard lists the limits of the account you are signed in to, with their reset times. OpenAI names it as the place to check them, and /status as the way to see the remaining limits during a CLI session.

/statusline chooses the items in the CLI footer, rate limits and context among them, and saves the choice in config.toml. /usage shows the account's token activity by day, by week or in total, and redeems an earned reset when one is available.

Codex usage dashboardOpenAI: where to see your limitsOpenAI: Codex CLI commands

Which limits apply?

Limits depend on the plan. Plus and Standard Business are metered per five-hour period; OpenAI says Pro currently has no five-hour limit. Weekly limits may also apply. Local messages and cloud chats share one allowance, and cloud tasks may use more of it than local messages.

The model, the context, reasoning, tool use, retrieval and caching all change how fast usage grows. Code Review usage applies only to reviews Codex runs through GitHub; a review run locally or outside GitHub counts toward your general limits.

Context is a different number

Codex also reports how much of the model's context window a conversation has filled. That belongs to one conversation; the limits are your plan's allowance. A conversation near full context can still have allowance left, and a conversation with room to spare can still reach a limit.

What Agent HUD shows

Every window Codex reports is a row, named by its own length, such as five hours or a week, with the share used, or the share left if you prefer, and the time it resets. Another limit Codex reports, such as one for a single model, appears under its own name. The glow turns yellow from 70% used and red from 90%.

The earned resets your account has available are counted beside the windows, and a new one raises an alert. Agent HUD never redeems them.

Agent HUD asks the Codex already on your Mac for the account's limits. It starts no turn and sends nothing to a model.

When the Codex app and the CLI use the same sign-in, they show one set of windows. A ChatGPT sign-in in Pi draws on the same plan and is counted once. More Codex accounts can be followed side by side, each read in turn, and the same email in two workspaces stays two accounts.

Codex's approvals and questions are answered on the HUD. Its approvals need Agent HUD's hook trusted in Codex.

Codex in Agent HUDApprovals, questions and plans on the HUD

Will it last until the reset?

The burn rate is the share of a window used per hour in its current cycle. If that pace empties the window before its reset, the row shows when it runs out; otherwise it shows the share you will have used by the reset.

Agent HUD alerts you once when a window is at risk, because the forecast says it runs out before its reset or because it crosses 90%. It alerts you again when the window is used up, and when a new reading confirms the reset.

Will your quota last until the reset?

Which session used the most?

Agent HUD's statistics count each Codex session's tokens with its turns, calls, context fill and cache hits, and every call with its model and the tools it asked for. The cost is an estimate at OpenAI's list price, marked ≈. It is not a charge.

Where did the tokens go?

When a window runs out

OpenAI says a turn already running when you reach a limit can continue. After that, Plus and Pro can buy credits to keep working, an API key can run more local chats, and /usage redeems an earned reset when your account has one.

Agent HUD shows when each window resets, and tells you when a new reading confirms the reset.

OpenAI: when you hit a usage limit