AI subscription usage and spend

How Limn uses your Claude or ChatGPT subscription, the weekly spend budget, and what automatic runs do when it is reached.

On this page

Limn has no AI service of its own. Every review, fix and summary runs on your own Claude or Codex account, from your computer, and counts against your own plan’s limits or your own API bill. This page explains which account Limn uses, how much of it Limn may use on its own, and what happens when that share runs out. The settings are on the AI tab, under AI spend.

#Which account Limn uses

Claude

Limn picks one Claude credential, in this order:

  1. Your Claude Code login on this computer (a Claude Pro, Max, Ultra, Team or Enterprise subscription). Run claude in a terminal and sign in. Runs then use your subscription’s usage.
  2. An Anthropic API key, set as ANTHROPIC_API_KEY in the environment Limn starts in. Runs are then billed by Anthropic at API prices.

When both exist, Limn uses the login and hides the key from the run, so a subscription run is never billed to the key by accident. To use the key first, tick Use my API keys first in Settings → AI (or start Limn with LIMN_AI_USE_API_KEY=true). Limn stores neither: it reads the login Claude Code already keeps, and the key only from the environment.

Codex

Codex runs only on a ChatGPT sign-in, made in Limn’s own Codex folder from the AI tab or with limn ai login codex. An OpenAI API key is refused for now, because Limn cannot yet show what its runs cost. Your own Codex setup and sign-in are never read or changed.

#Your plan’s limits are shared

On a subscription, Limn’s runs come out of the same limits as your own use of Claude or ChatGPT. Anthropic says so for Claude:

These limits are shared across Claude and Claude Code, meaning all activity in both tools counts against the same limits.
Use Claude Code with your Pro, Max, or Ultra plan | Claude Help Center (retrieved 2026-11-09)

Claude limits each week, and on some plans each five-hour session too. Limn’s budget is a share of the week, and of the session where your plan has one.

#The spend budget

Limn runs some AI work by itself: auto review and re-reviews, Check threads, auto fix, the CI and ticket reviews, diagrams, This Sprint’s briefs and the Slack summary. The spend budget limits only those automatic runs.

Illustration: AI spend on a Claude subscription, at the default share.

On a subscription

  • The budget is on by default at 50% of your week. Change the share, or switch the budget off.
  • If your plan limits each five-hour session, Limn takes the same share of it (50% of the week means 50% of each session). Open Set a separate 5-hour limit to choose a different share; Use the weekly share goes back. Anthropic has removed the five-hour limit from some plans. Limn shows the five-hour bar and setting only when your provider has reported a five-hour limit in the last 12 hours.
  • Everything is shown in percentages of your plan, never in dollars.
  • Claude: Limn adds up what its own runs use and compares it with your subscription’s usual weekly allowance. Anthropic does not publish exact figures, so this is an estimate. Choose Your subscription (Pro, Max 5x, Ultra or Team) if Limn asks; when Claude reports only “Max”, Limn asks whether it is Max 5x or Ultra rather than guess.
  • Codex: ChatGPT reports how much of its weekly limit, and five-hour limit where there is one, your account has used, in whole percent. Limn’s share is that figure times the part of your Codex use that came from its own runs, so it moves in coarse steps.

On an API key

There is no limit until you set one. Select Set a budget and enter a Weekly budget (US$). The section then shows what Limn’s runs have cost this week, in US dollars.

#When Limn’s share is used

  • Automatic runs are skipped, not queued. A message says, for example, Limn has used its share of your week. Automatic runs resume Tue 09:00. They start again when the week or the session resets.
  • Your own runs are never refused. Over budget, a Run button asks for a second press to confirm, then runs.
  • Your own use of Claude or ChatGPT outside Limn is never limited by Limn.
  1. Open Settings, then the AI tab.
  2. Under AI spend, tick or untick the budget, and type Limn’s share.
  3. Select Save.

#Effort and how much a run uses

Higher Effort reads more and finds more, and uses more of your plan. Turning off design diagrams saves about a third of a review’s usage. A Claude + Codex review uses both providers’ limits. Each run shows its cost: a share of your week on a subscription, or US$ on an API key. For the history of your use, see AI usage.

#Heavy or automatic use

Anthropic sets its subscription limits for one person’s ordinary use:

Advertised usage limits for Pro, Max, and Ultra plans assume ordinary, individual usage of Claude Code and the Agent SDK.
Legal and compliance - Claude Code Docs (retrieved 2026-11-09)