Plan capacity
Set and reset estimated weekly and monthly API-dollar capacity.
Plan capacity is at the bottom of each agent's page in Settings → Agents, and behind the Plan capacity settings button in Usage history. Expand Plan capacity (advanced) to edit API-priced usage per 1% of the weekly window, or the Weekly or Monthly dollars it implies (per 1% × 100). Changes save on blur on this device. Reset removes your override and returns to a locally learned estimate, or the default estimate when there is not enough history.
Usage history's charts don't use your saved override, learned values or measured ceilings yet; their capacity estimates use the plan or harness default. Cost comments on pull requests already use them.
These figures estimate the API list-price value of included usage. They are not subscription prices, credits, invoices, or guaranteed provider allowances. Changing them does not change provider limits or reset times. Monthly is weekly multiplied by 30.4 / 7, not a separate monthly quota.
Usage percentages track your account's allowance, not the provider's load. A message such as Selected model is at capacity means the provider is overloaded and can appear at any usage level; see Common errors.
Where the defaults come from
Capacity uses your override, then a learned per-1% value, then a measured ceiling from recorded spend at a weekly limit, then a plan or harness default. The learned value is the API-priced spend since the weekly reset (or up to a limit hit) divided by the percentage points the reading rose; it needs at least 10 steps. Settings also shows Latest 1% and Observed window, each with its step count and its gap from the default. Missing capacity does not mean zero. Older unqualified Claude “Max” assumes Max 5x; unqualified Codex “Pro” assumes Pro $100. These assumptions do not identify the account's actual tier.
| Plan | Weekly estimate | Basis |
|---|---|---|
| Claude Pro / unknown Claude tier | $170 | Wokey's own measurements, 559 accounts, Oct 5–12, 2026. |
| Claude Max 5x / unqualified Max | $500 | Rounded from Botfarm's $523 measurement, February 2026; an older workload and lower-bound estimate. |
| Claude Max 20x | $1,750 | Wokey's measurement of 36 accounts, rounded from $1,751. |
| Codex Plus / unknown Codex tier | $66 (~$0.66 per 1%) | Estimate, close to InventiveHQ's $66 projection, Oct 4, 2026. |
| Codex Prolite / Pro $100 / unqualified Pro | $354 (~$3.54 per 1%) | Estimate, within InventiveHQ's own measurements of $330–380 across two windows. |
| Codex Pro $200 | $655 (~$6.55 per 1%) | Estimate for new subscribers, within InventiveHQ's $640–670. Until Oct 29, 2026 the default is the grandfathered ~$13 per 1% ($1,300). |
| Codex Pro $500 | $1,640 (~$16.40 per 1%) | Estimate. |
| Cursor Pro / Pro+ / Ultra | $5 / $14 / $46 | Explicit fee proxy: legacy monthly subscription prices of $20 / $60 / $200 × 7 / 30.4, rounded. This is a weighting assumption, not the included API allowance. Unknown Cursor tiers use $5. |
| Ollama Cloud Pro / Max | $14 / $69 | Ollama's included monthly credits of $60 / $300 × 7 / 30.4, rounded, at Ollama's rates. Unqualified Ollama Cloud assumes Pro; Free credits are unspecified. |
| Grok / Kimi / Antigravity / Devin / unlisted plans | $65 | Neutral weighting assumption using the Codex Plus seed above. These accounts have no calibrated API-dollar default; this does not imply equal provider allowances. |
Grok, Kimi, Antigravity and Devin describe their plans and limits without an API-dollar weekly pool usable here. Their shared fallback avoids inventing different dollar amounts without evidence. New Kimi plans have no weekly window; older plans may retain one. Ollama's current credit plans differ from older session/weekly plans. A weekly equivalent is a comparison weight, not a reset schedule.
Defaults weight each account with a usable plan-limit reading in the Usage Overall capacity row and fill Settings fields. A reading counts until its window resets; after the reset the account counts as unused. An account with a known plan but no reading yet counts as unused too, marked as an estimate. Accounts with no plan or limit (free, local inference, API credits only) are left out of the calculation and its coverage count. Exhausted accounts contribute their full weight with nothing remaining. These defaults do not replace recorded percentages, measured ceilings, model-specific capacities, or historical graphs.
Sources checked October 17, 2026. Published measurements describe their authors' workloads, not provider guarantees. Models, cache rates, missing local activity, plan changes and account-specific limits can change the comparison substantially. Default estimate, Learned estimate, and Your override identify the basis. Prefer your own history or override when a default does not fit.
Which accounts appear
Plan capacity and the Usage account count omit removed accounts and identities currently signed out. An exhausted account is still signed in and remains visible. Removing an account does not erase its saved usage, snapshots, or overrides; historical spend remains in the Usage totals.