Inter-office memo
- To
- Everyone whose Claude Code limits run out too soon
- From
- Mo, Chief Executive
- Date
- November 15, 2026, 10:45 PM CST
- Re
- Making Claude Code usage last
How to make Claude Code’s usage limits last longer in 2026: 9 ways compared
Start with /usage, which shows what’s using your Claude Code limits and suggests fixes. For most people the big savings are clearing between tasks, Sonnet instead of the default Opus for routine work, and lower effort for small changes. Heavy users should rein in subagents and trim CLAUDE.md. Codex or a local model can take routine work, and Enjoy hands it over mid-conversation.
Disclosure: Enjoy is our product. We checked every command, limit and price below against Anthropic’s, OpenAI’s, OpenCode’s and Ollama’s own pages on November 15, 2026, and we say where another way is the better pick, including the free ones that don’t need Enjoy.
The ways to make Claude Code’s limits last, compared
Claude Code doesn’t count messages. “How fast you reach a limit depends on the model you use, how long your chats are, and how large your files are,” says Claude’s pricing page, and Anthropic’s help center adds “the features you use” and “the effort level you’ve selected.” Pro and Max have a five-hour session limit and a weekly limit, and they’re “shared across Claude and Claude Code, meaning all activity in both tools counts against the same limits,” Anthropic says. On Pro, Max, Ultra and Team, Claude Code starts on Opus 5.5 at medium effort, per its model docs.
| Way | Cost, checked November 15, 2026 | What it cuts | Best for |
|---|---|---|---|
| Find what’s using it, with /usage | Free | Nothing by itself: it names what to change | Not knowing where your usage went |
| Clear between tasks | Free | History sent again with every message | Sessions that run all day or wander |
| Sonnet for routine work | Free; Fable on Pro is paid from usage credits | Opus’s higher cost per turn | Features, tests, known bugs |
| Lower the effort | Free | Thinking, billed as output | Small, clear changes you review |
| Trim what loads every turn | Free | CLAUDE.md, MCP listings, pasted files, long command output | A long CLAUDE.md or many MCP servers |
| Rein in extra agents and background work | Free | Requests from subagents, teams, workflows and idle sessions | Heavy use of subagents, ultracode and /loop |
| Plan first, and be specific | Free | Broad scans and redone work | Changes across more than a few files |
| Codex or a local model for routine work | Codex: Free, Go $8 a month, Plus $20, Pro from $100; a local model is free | Work moves off Claude’s limits | Work that doesn’t need Claude |
| Enjoy | A monthly subscription, free for 7 days, with your phone and one computer (pricing) | Shows each conversation’s size, and moves work to another agent mid-conversation | Seeing and spreading usage across agents |
Prices are each company’s US list prices, before tax. The first seven ways are settings and habits in Claude Code itself, and Codex on ChatGPT’s Free plan and a local model cost nothing extra either.
Why Claude Code runs out so fast
Claude Code “sends your full conversation with every request, and each time Claude uses tools it sends another request carrying that batch of tool results,” says its cost guide. Prompt caching makes the repeats cheaper, but “a one-line question in a session that has been open all day still draws usage for the whole conversation.” Anthropic’s guide to models and limits in Claude Code puts it bluntly: “Almost every ‘I burned through my limit by lunchtime’ report traces back to one of these five” habits: clearing between tasks, matching the model to the job, pointing at files instead of pasting them, a lean CLAUDE.md, and a plan before big changes. The ways below cover those five, and a few more.
Find what’s using it, with /usage
Best for: anyone who doesn’t know where the usage went. Start here.
On a Pro, Max, Ultra, Team or Enterprise plan, /usage shows “plan usage bars, activity stats, and a usage breakdown,” says Claude Code’s cost guide. The breakdown attributes recent usage “to skills, subagents, plugins, and individual MCP servers, each shown as a percentage of the total,” then lists a recommendation for anything behind at least 10% of it, such as 25% Trim ~/code/api/CLAUDE.md · 49k tokens per request. Press d or w to switch between the last day and the last week. When the footer shows a hint such as f to fix 2 of 3, press f: Claude Code runs any /effort change itself, and Claude shows you each of the other changes it can make before making it. To see what fills one conversation, run /context.
Price: free. The f shortcut needs Claude Code v2.1.305 or later.
Watch for: “The figures are approximate and computed from local session history on this machine, so usage from other devices or claude.ai is not included.” And a row’s share “tells you how much of your usage the row applies to, not how much the change saves.”
Clear between tasks, and compact when you carry on
Best for: sessions that run all day, or wander from one problem to the next.
The biggest cost is history. “Every previous message is resent on every turn, so a session that has wandered through three unrelated problems pays for all three on each new message,” says Anthropic’s help center. Its rule of thumb: “use /clear when starting a new task, and /compact when continuing a long one,” and its test: “if your next prompt would make perfect sense in a brand-new session, clear before sending it.” Run /rename first to find the session later with /resume, and tell /compact what to keep, as in /compact Focus on code samples and API usage.
Breaks count too. On a Claude plan, the main conversation’s cache lasts an hour, per Claude Code’s caching docs, and its cost guide warns that “your first message after a break longer than the cache lifetime misses the cache and reprocesses your full context.” For a conversation of 200,000 tokens or more that you’ve left for about an hour, Claude Code’s idle compaction summarizes it first, so “Your next message then sends a summary instead of the whole conversation.”
Price: free.
Watch for: “/clear cannot be undone,” so copy out anything you still need. And compacting isn’t free: “/compact reads the conversation it summarizes, so compacting a large context is itself a large request. When you want a fresh start instead of continuity, /clear costs nothing.”
Use Sonnet for routine work
Best for: most day-to-day coding: features, tests, known bugs and refactors.
“Opus costs several times more per turn than Sonnet, and Sonnet more than Haiku. Spending Opus on routine work is the fastest way to drain a daily limit,” says Anthropic’s help center. Its defaults: Sonnet “for most coding,” Opus “when you’re genuinely stuck or the change is wide,” and Haiku “for quick mechanical work.” Run /model sonnet, or set a default in /config. To plan on Opus and build on Sonnet, /model opusplan “uses opus during plan mode, then switches to sonnet for execution,” per the model docs. For simple subagent tasks, the cost guide suggests model: haiku in the subagent’s settings.
Price: free. Fable is the exception: Claude’s pricing lists it on Pro as “paid with usage credits,” and on Max as “included, up to half your weekly limit.”
Watch for: switch when a task starts, not halfway through. “Each model has its own cache,” says the caching page, so after /model, “the next request reads the entire conversation history with no cache hits,” and with opusplan, “each plan-mode toggle is a model switch and starts a fresh cache.” Switching doesn’t rescue you after a session or weekly limit either: those are “shared across all models.”
Lower the effort
Best for: small, clear changes you’ll review yourself.
Effort sets how much Claude thinks, and “Thinking tokens are billed as output tokens,” says the cost guide. Fable 5.2, Opus 5.5, Sonnet 5.5 and Haiku 5.5 default to medium, and the model docs suggest low for “Quick exchanges where you review each result, such as brainstorming, a first sketch, or a small change like a rename.” Run /effort low, or press s in the /effort slider to change it for this session only. If you raised it to max, bring it back down: it “may show diminishing returns and is prone to overthinking.” For one hard turn, type ultrathink in your prompt, which asks for “deeper reasoning on that turn without changing your session effort setting.” On Opus 5.5 and Sonnet 5.5 with a Claude plan, changing effort keeps the cache, so it’s safe mid-task.
Price: free.
Watch for: less thinking means less checking. In Anthropic’s tests, “At a lower level, Claude returned a starting point sooner,” while at a higher one it “tested more edge cases and verified more of its work.” And the Ultracode toggle in the same slider works the other way: “a session with ultracode on reaches a session or weekly limit sooner than the same work with it off,” says the workflows page.
Trim what loads with every message
Best for: projects with a long CLAUDE.md, several MCP servers, or big pasted files and logs.
Some context rides along on every turn. CLAUDE.md “is loaded into context at session start,” and the cost guide says to “keep CLAUDE.md under 200 lines by including only essentials,” moving instructions for particular workflows into skills, which “load on-demand only when invoked.” Splitting it with imports doesn’t help, since imported files “load together with the CLAUDE.md that imports them,” per the memory docs.
MCP tool definitions are now deferred until Claude uses one, but “Tools like gh, aws, gcloud, and sentry-cli are still more context-efficient than MCP servers,” and /mcp turns off servers you aren’t using. For files, Anthropic’s help center says to name a path instead of pasting: “the @ prefix injects the entire file plus its CLAUDE.md tree into context, so use a bare path when you are trying to save tokens.” And a hook can cut a command’s output before Claude reads it: the cost guide’s example keeps “only the last 100 lines” of test output.
Price: free.
Watch for: caching already makes repeated context cheaper. Anthropic notes that turns after the first are billed “at the much cheaper cache-read rate, so the dollar cost is lower than the raw line count suggests,” so trim what /usage or /context flags as large first. Don’t trim what saves work, either: a skill describing your project’s layout spares Claude “spending tokens reading multiple files to understand the structure.”
Rein in extra agents and background work
Best for: heavy users of subagents, agent teams, ultracode, /loop and goals.
Every agent sends requests of its own. A subagent “sends its own requests, which count toward the same usage limits as your main conversation,” says the subagents page, and the cost guide says “Agent teams use approximately 7x more tokens than standard sessions when teammates run in plan mode.” A dynamic workflow “spawns many agents, so a single run can use meaningfully more tokens than working through the same task in conversation,” per the workflows page. The /usage breakdown shows each one’s share, and its Loops rows list your heaviest scheduled tasks.
Some usage happens while you’re away. In an idle session, a scheduled task, a message from another of your sessions, or a goal check-in each starts a turn that sends your full context, the cost guide says. Setting crossSessionInbound to hold holds those messages, and CLAUDE_CODE_GOAL_CHECKIN_MINUTES=0 turns check-ins off. Prompt suggestions are “mostly cache reads plus a few output tokens,” and Claude Code already skips them “while your account is close to or at its usage limit.” You can turn them and session recaps off in /config.
Price: free.
Watch for: subagents can save context as well as spend it. The cost guide suggests handing “Running tests, fetching documentation, or processing log files” to one “so the verbose output stays in the subagent’s context while only a summary returns.” Keep subagents for that, on a smaller model where the job allows. Our memo on Anthropic’s advice covers when one agent with skills does better.
Plan first, and be specific
Best for: changes that touch more than two or three files.
Wrong turns are expensive. “A plan costs a few hundred tokens. A wrong 400-line diff that you revert and regenerate costs thousands, twice,” says Anthropic’s help center, which suggests plan mode “for anything touching more than two or three files.” Press Shift+Tab to cycle to it, and “Claude explores the codebase and proposes an approach for your approval,” says the cost guide. It adds: “Vague requests like ‘improve this codebase’ trigger broad scanning. Specific requests like ‘add input validation to the login function in auth.ts’ let Claude work efficiently with minimal file reads.” If Claude heads the wrong way, press Escape, and /rewind restores “conversation and code to a previous checkpoint.” A test it can run lets Claude catch “issues before you need to request fixes.”
Price: free.
Watch for: planning runs on your plan too, and on Opus with opusplan. For a one-line fix, skip it and say exactly what you want.
Hand routine work to Codex or a local model
Best for: work that doesn’t need Claude, so Claude’s limits go to work that does.
Other agents have limits of their own. OpenAI’s Codex is in every ChatGPT plan from Free, per Codex’s pricing, with GPT-6.1 Sol from Plus. For no limit at all, OpenCode runs local models: “Start Ollama and pull a model. OpenCode probes http://127.0.0.1:11434, discovers completion models, and adds them to /models without an account connection,” say its docs. One built for this is Recursive’s Flint, “a 39.8 billion-parameter mixture-of-experts model from Recursive, built for coding agents,” which ollama run flint downloads. Our guides to using Claude Code and Codex together and running a coding agent on a local model compare the ways to split the work.
Price: Codex is $0 on ChatGPT Free, $8 a month on Go, $20 on Plus, and Pro is “Plans at $100, $200, or $500 USD per month.” A local model is free, on your own computer.
Watch for: each starts a conversation of its own, so tell it where things stand. Codex’s Free and Go plans get “GPT-6 Luna at Standard speed in the desktop app, subject to rollout,” and from December, OpenAI says, “Plus, Pro, and Business plans’ usage allowance will be shared across ChatGPT and Codex.” For local models, “OpenCode only adds models whose Ollama metadata includes the completion capability.”
Enjoy
Best for: seeing where each conversation’s usage goes, and moving routine work to another agent without starting over.
Enjoy runs Claude Code, Codex, Grok Build, Antigravity, Muse Code, and local models through OpenCode, on Mac, Windows and Linux, on the plans you already have. For usage, it adds:
- Each conversation’s size. A Claude Code conversation’s ⋯ menu shows the tokens it has used, and Usage opens every number with what it means, including how large the conversation has grown. Show in conversations keeps how full its context window is beside its ⋯ button, with Compact now… a click away (how).
- Every account’s limits. The agent settings show how much of each limit is left and when it resets, for each signed-in account (how).
- Another agent, mid-conversation. Switch to Codex or a local model for the routine part, and “The new agent gets the conversation so far.” Switch back for the hard part, and Claude Code “continues where it left off” (how).
- Model and effort per message. The agent settings’ Model and Effort columns change them for your next message, and our help guide to making usage last covers the habits above.
When a limit arrives anyway, Enjoy moves the conversation to another of your accounts or waits for the reset, as our guide to what to do when Claude Code hits its limit explains.
Price: Enjoy is one monthly subscription, and the first seven days are free, with a card to start. It includes your phone and any browser and one computer; each other computer costs extra each month, and each teammate you invite needs a subscription of their own. See pricing.
Watch for: Enjoy doesn’t change Anthropic’s limits or make Claude Code use fewer tokens on its own: the settings above do that. Per-conversation usage is for Claude Code only, in the app on your computer, since “Other agents don’t report what each turn used yet.” An agent you switch to reads the conversation so far on its own plan. Enjoy isn’t open source, and it’s a monthly subscription on top of your AI plans once the free week ends.
When to stay with the defaults, or upgrade
Opus 5.5 at medium is the default for a reason: in Anthropic’s testing, “Opus 5.5 at medium matches or exceeds Opus 5 at high on coding and knowledge-work evaluations,” per the model docs. Of Fable, Claude Code’s most capable models, the docs say “root-cause investigations, outage debugging, and architecture decisions are where the extra investigation and verification pay off.” Keep Opus, or Fable, for the work where you’re stuck, and save on the routine work around it.
If you’ve done all this and still run out most weeks, the plan is too small. Claude’s pricing lists Max at $100 a month for “5x Pro usage every 5 hours and every week,” and Ultra at $200 for “10x Pro usage every week,” with no five-hour limit for verified accounts, against Pro’s $20, or $17 a month billed yearly. Our guide to Claude Pro vs Max vs Ultra helps you pick.
How we compared
We read Claude Code’s cost, model, prompt caching, memory, MCP, subagent, workflow, interactive mode and settings docs, Anthropic’s help center and pricing, OpenAI’s Codex pricing and help center, and OpenCode’s and Ollama’s pages on November 15, 2026, and every one is linked below. We quote each company’s own words for its limits, and leave out what it hasn’t documented, including other sites’ estimates of how much each habit saves. Plans and limits change often, so we recheck this guide against its sources at least once a month, and the date at the foot of the page says when we last did.
14.Questions.
Why does Claude Code use up my limit so fast?
- Mostly because it sends the whole conversation with every request. Claude Code’s docs say “a one-line question in a session that has been open all day still draws usage for the whole conversation.” Opus, the default model, higher effort, subagents and a long CLAUDE.md add to it. Run /usage to see what takes the most.
Does Sonnet use less of my Claude Code limit than Opus?
- Yes. Anthropic says “Opus costs several times more per turn than Sonnet, and Sonnet more than Haiku.” Switch with /model sonnet at the start of a task, since a switch mid-conversation reads the whole history again without the cache.
Does /compact save usage?
- For the messages after it, yes, but compacting is a large request of its own, because it “reads the conversation it summarizes.” When you’re starting something unrelated, /clear costs nothing.
Do chats in the Claude app count against my Claude Code limit?
- Yes. Anthropic says the limits are shared across Claude and Claude Code, “meaning all activity in both tools counts against the same limits.” The /usage breakdown in Claude Code covers only the computer you run it on, so it won’t show what you used on claude.ai.
Does lower effort make Claude Code worse?
- On hard work, it can. At a lower level, Anthropic found Claude “returned a starting point sooner,” while at a higher one it “tested more edge cases and verified more of its work.” Use low for small changes you review, and keep medium, the default, or higher for bugs and wide changes.
Can another agent take the routine work?
- Yes. Codex on a ChatGPT plan has limits of its own, and a local model through OpenCode has none. Enjoy moves a conversation to one of them with the conversation so far, and back to Claude Code for the hard part.