Guides to AI coding tool limits and spend
What to do when you run out of Claude Code or Codex allowance, whoever pays for it, and why it happened. How the limits on Claude Code, Codex, Cursor and OpenCode work, how to see where you stand, how to see what a team spends, and how the tools that track it compare. Each page answers one question, is dated, and lists the vendor pages it was checked against.
You've just run out
- You've hit your Claude Code or Codex limit: what you can do nowThree choices when Claude Code or Codex stops you: wait for the window to come back, and when it does; keep working by switching model, paying for usage past the limit or moving to the other tool; or make the next window last longer. Plus what to do if your company pays. Checked against Anthropic's and OpenAI's pages.Checked on 5 November 2026
- Keep coding when Claude Code or Codex cuts you off: hand the task to the other toolIf you pay for both, ask the agent you are leaving to write a short HANDOFF.md, then have the other tool read it and carry on. Tried both ways on real tasks: the second tool finished the work from the note alone. The two prompts, what happened, what it costs with the arithmetic shown, and what to watch.Checked on 6 November 2026
- Find out what ate your Claude Code or Codex window: a prompt for your own agentPaste one prompt into Claude Code or Codex and your own agent reads the session logs on your machine and reports the sessions and requests that used the most, the re-reads after a break, and how far each Codex session moved your windows. Nothing is sent to OpenReason. Tested on real logs in both tools. If your window is full, the free Session Forensics page does a similar job in your browser with no allowance.Checked on 6 November 2026
How the limits work
- How Claude Code's 5-hour and weekly limits workHow Claude Code's 5-hour and weekly limits work: the 5-hour window opens on your first message and resets in full five hours later, the weekly limit resets at a fixed time assigned to your account, both are shared with claude.ai, and verified Ultra accounts have no 5-hour limit. What counts against them and how the plans compare, from Anthropic's pages.Checked on 5 November 2026
- How Codex's 5-hour and weekly limits workHow Codex's 5-hour and weekly limits work: Plus and Business Standard seats have a 5-hour window and Pro plans do not, a weekly limit applies on top, local messages and cloud tasks share one allowance, and from December ChatGPT and Codex share one allowance. With OpenAI's own wording and dates.Checked on 5 November 2026
- Why did my Claude Code limit run out so fast?Usually because of context, not how much you typed: a long session sends the whole conversation with every request, a break past the cache lifetime makes the next message cost full price, and subagents, compaction and big Opus sessions add up. What each kind of work costs against the 5-hour window, what to do now, how to find the cause with /usage, what people have reported, and what differs on Codex.Checked on 5 November 2026
- How to check your Claude Code usage, and see it over timeJust ran out? Run /usage in Claude Code for your 5-hour and weekly bars and what has been using them, or open Settings > Usage on claude.ai. Put the percentages in your status line with /statusline, and keep a record across machines with ccusage, claude-session-budget or OpenReason Intel. What each one reads and what it cannot tell you.Checked on 5 November 2026
- How to check your Codex usageJust ran out? Type /status in the Codex CLI for each window's share left and reset time, or open the usage dashboard at chatgpt.com/codex/settings/usage. Codex writes each limits snapshot to its session file, which is how OpenReason Intel keeps a record of your 5-hour and weekly windows across machines.Checked on 5 November 2026
- How Cursor's usage limits and on-demand pricing work, and how to check your Cursor usageCursor's included usage is a pool for its own models that refreshes weekly on The X Subscription and monthly on legacy plans; third-party models are paid from prepaid credit at API list prices, capped by a monthly limit. Where individuals and team admins check usage. Checked against Cursor's pages.Checked on 4 November 2026
- How OpenCode's limits work, when it runs on your own provider plans or keysOpenCode has no limits of its own: the limit is your provider's. An API key has no 5-hour or weekly window; a ChatGPT Plus or Pro sign-in counts against your ChatGPT plan; Claude Pro and Max are not a supported route; OpenCode Go has its own 5-hour, weekly and monthly ceilings. Checked against OpenCode's, Anthropic's and OpenAI's pages.Checked on 4 November 2026
Make it last
- What people do when they run out of Claude Code or CodexMost wait. The rest switch tools, start the window early so the reset lands mid-day, watch the meter and stop near the end, start fresh instead of compacting, use a cheaper model for routine steps, upgrade, or ask their company for more. Each in people's own words from Reddit, with what it costs and what you get.Checked on 5 November 2026
- How to make your Claude Code limit last longer: the tools that helpClaude Code's own /usage (with its fix shortcut), /context and status line first, then Caveman, ccusage and tokscale, Session Forensics, a handoff note and OpenReason Intel. What each does to make a Claude Code or Codex allowance last, what it does not, and its own numbers, read from each tool's page.Checked on 6 November 2026
- ccusage, tokscale, UsageAtlas, CodeBurn and the other local trackers vs OpenReason IntelA fair comparison of nine local AI coding usage trackers (ccusage, tokscale, UsageAtlas, CodeBurn, OpenUsage, Usagebar, claude-session-budget, ClaudeCapper, Claude Code Usage Monitor) and OpenReason's free Session Forensics page with OpenReason Intel, from each tool's own pages: when a local tracker is enough for one person, what Intel adds for one person and for a team, which tools each covers, whether it notifies or stops anything, and the price.Checked on 4 November 2026
For teams and budgets
- How to ask your company for more Claude Code budgetRead the message to see whose limit ran out: your seat's 5-hour or weekly allowance, or a spend limit your company set. On Team and Enterprise, /usage-credits sends your admins a request. What approvers say they want, a five-line request, and what people say the process is like.Checked on 5 November 2026
- What happens when your team's Claude credits run outWhen a Claude Team or Enterprise organisation reaches its monthly spend limit on usage credits, members who have used up their seat stop paid usage until an Owner raises it or the month resets; each seat's own window still resets. The Team plan's API credits never apply to Claude Code. What to do as the engineer, and what to change as the admin.Checked on 5 November 2026
- How to see what your team spends on Claude Code, and who is using itWhat a Team or Enterprise Owner can see in Anthropic's analytics and spend report, what the Console shows on the API, why a team on Max plans has no view at all, and how OpenReason Intel shows seat commitments and variable spending together per person.Checked on 3 November 2026
- How to find out who on your team went past their plan, and what it costOn a Claude Team plan nobody can spend past their seat unless usage credits are on; then the Owner sees each member's month-to-date spend and sets caps. Max plans have no team view. OpenReason Intel shows each over-limit period with the session, repo, model and turns around it.Checked on 3 November 2026
- How to see Claude Code, Codex and Cursor spend in one place, and set a budget across a teamNo vendor's dashboard shows another vendor's tool. What Anthropic, OpenAI and Cursor each show a team admin and what they leave out, how OpenReason Intel shows Claude Code and Codex seats and variable spending together per person, and how to set a budget across tools.Checked on 4 November 2026
- Claude Team vs Max vs API for an engineering team: which costs less?Every Claude plan's price and allowance from Anthropic's own pages, with the sums for five engineers on Team Standard, Team Premium, Max, Ultra and the API. Standard seats are cheapest at $20 a month; a Premium seat and Max both cost $100, and the Premium seat gives 6.25 times Pro per session against Max's 5 times.Checked on 4 November 2026
- Tools that track AI coding spend across a team: TokenShift, Vantage, CloudZero, Faros AI, Jellyfish, Weave, Requesty, Torii and OpenReason IntelEight products people ask about for tracking a team's spend on Claude Code, Codex, Cursor and other coding agents, plus OpenReason Intel, each described from its own pages: where its figures come from, which tools it covers, whether it sets budgets, and the price.Checked on 4 November 2026
What changed
- Claude Code and Codex limit changes: a dated logEvery change to Claude's and OpenAI's coding limits we have confirmed on the vendor's own page, newest first, with the date, the plans it affects and the source: Ultra's 5-hour limit removed on 30 October 2026, ChatGPT and Codex sharing one allowance from December, and Codex's 5-hour limit lifted in July and back for Plus in August. User reports are labelled as reports. Get an email when the next change is confirmed.Checked on 5 November 2026
- Is the Claude 5-hour limit gone?Only for verified Ultra accounts, the $200 plan, since 30 October 2026. Pro and the $100 Max plan still have the 5-hour limit, as do Team and Enterprise seats, and every plan keeps a weekly limit. How to check your own account, and where Codex stands.Checked on 5 November 2026