The Claude Fable 5 Operator's Guide
Anthropic's most powerful model just landed in your Max plan. Most people will spend it like a chatbot and wonder where their weekly limit went. This page is the system for getting real work out of it.
Stop prompting. Start delegating.
TLDR
Claude Fable 5 is now included in paid Claude plans, and it is not a chatbot upgrade. It plans before it executes, checks its own work, puts helpers on big jobs, and keeps going for hours. The catch: it drinks your weekly limits if you feed it small talk. The 12 prompts below, adapted from Anthropic's own guidance, show it how you want to work: proper briefs, honest progress reports, overnight runs that actually finish. Everything is free to copy on this page, and the whole set downloads as one markdown file for your prompt library.
What actually changed with Fable 5?
Every other model answers your question. Fable 5 runs your workflow. Give it a job and it plans first, breaks the work down, spins up helper agents for the independent pieces, checks the output against what you asked for, and keeps going for hours without hand-holding.
That changes how you prompt. You are no longer writing clever one-liners for a clever autocomplete. You are briefing a capable operator who will make hundreds of decisions while you are not watching. The quality of the brief decides the quality of the week.
The facts, so you do not have to sort them from the hype: Anthropic announced Claude Fable 5 on 9 June 2026, access was restored on 1 July 2026 after a gap Anthropic has not fully explained, and it is now available to Pro, Max, Team and Enterprise users inside their existing plan limits.
The old way: one assistant drowning in your backlog. The Fable 5 way: the job split across a team that researches, builds, tests and reviews.
When should you NOT use Fable 5?
This is the section that saves your credits. Fable 5 works longer and harder per request than any other model, which is exactly what you want on big jobs and exactly what you do not want on small ones.
Keep this on Opus or Sonnet
Send this to Fable 5
The one-line rule: if you would hand the job to a capable contractor with a written brief, it is Fable work. If you would ask a colleague across the desk, it is not.
What do you want Fable 5 to do?
Twelve prompts, adapted from Anthropic's official Fable 5 guidance and translated for people who run businesses rather than write code. Jump straight to the one you need.
What should you give it before any prompt?
Your files. Context is king, and every prompt on this page works ten times harder when Fable 5 can see what you see. Before you hit send, upload the plans, the drafts, the price lists, the contracts, the meeting notes, the spreadsheet you have been avoiding. Messy is fine. It reads faster than you do.
The pattern for every big job: attach everything relevant, then brief it, then let it run. A rough brief with the real files beats a perfect brief with none.
How do you brief Fable 5 so it nails the job first time?
Fable 5 makes hundreds of small decisions during a long job. If it does not know who the work is for, it guesses.
When to use it: Start every meaningful task with this, and attach the files that matter before you hit send: plans, drafts, price lists, contracts, meeting notes. Context is king, and Fable 5 can only work with what it can see.
I'm working on [the larger task] for [who it's for]. They need [what the output enables]. With that in mind: [request].
I'm working on a price review for my joinery business. It's for me and my ops manager. We need to see exactly where our margin has slipped so we can reset our rates before we quote the spring season. With that in mind: go through the three supplier price lists and last year's quotes in this project, compare them against our current rate card, and give me the 15 products where margin has slipped the most, with a recommended new rate and one line of reasoning for each.
Fable 5 checks every small decision against the goal you give it, so one line of context beats three paragraphs of instructions.
Download .mdHow do you stop Fable 5 overthinking a simple ask?
On smaller tasks Fable 5 can research and deliberate for minutes before it lifts a finger. That time comes out of your limits.
When to use it: Paste at the top of any task where you already know what you want and just need it done.
When you have enough information to act, act. Do not re-derive facts already established in the conversation, re-litigate a decision I have already made, or narrate options you will not pursue. If you are weighing a choice, give a recommendation, not an exhaustive survey.
This is Anthropic's own instruction for keeping the model moving when a task is clear enough to start.
Download .mdHow do you hand over a big messy job without a perfect brief?
Most people under-use Fable 5 by giving it small tasks. Its real strength is jobs you would normally break up yourself.
When to use it: Use when the job is bigger than one sitting and you have not fully thought it through. Let the model do the thinking.
Here is the job: [describe the outcome you want, even roughly]. Before you start, scope it. Tell me how you would break it down, what you need from me, and what could go wrong. Ask me your clarifying questions now, in one batch. Once I answer, execute end to end without checking in unless you are genuinely blocked.
Here is the job: I want a complete onboarding pack for new hires at my electrical contracting company. Rough shape in my head: a first-day checklist, a first-week training plan, the safety basics, and a who-to-call list. I have uploaded our current staff handbook, two old checklists and our org chart; keep whatever is worth keeping. It is not fully thought through, which is why I am handing it to you. Before you start, scope it. Tell me how you would break it down, what you need from me, and what could go wrong. Ask me your clarifying questions now, in one batch. Once I answer, execute end to end without checking in unless you are genuinely blocked.
Anthropic's advice is to start at the top of your difficulty range: pick a harder task than you would give other models, and have Fable 5 scope it before executing.
Download .mdHow do you get Fable 5 to check its own work as it goes?
A long run that drifts off spec in hour one wastes everything after it. Fable 5 can catch this itself, but only if you ask.
When to use it: Add to any long-running job: a build, a big document, a research project, a batch of pages.
Establish a method for checking your own work at an interval of [every major step / every 30 minutes / every section] as you build. Run this check every interval, verifying your work with subagents against the specification. If a check fails, fix it before moving on.
Anthropic found that fresh-context checkers outperform the model critiquing itself, like a builder who stops every hour to check the level.
Download .mdHow do you know its progress reports are actually true?
On long runs, any AI can drift into reporting what it intended to do rather than what it did.
When to use it: Add to every long or unattended run. Pairs perfectly with The Overnight Run below.
Before reporting progress, audit each claim against a real result from this session. Only report work you can point to evidence for; if something is not yet verified, say so explicitly. Report outcomes faithfully: if something failed, say so with the output; if a step was skipped, say that; when something is done and verified, state it plainly without hedging.
In Anthropic's testing this instruction nearly eliminated fabricated status reports, even on tasks designed to provoke them.
Download .mdHow do you stop it doing things you never asked for?
Fable 5 is proactive. Sometimes that means it drafts the email, changes the file, or fixes the thing when you only wanted an opinion.
When to use it: Use when you want analysis and advice, not action. Especially anywhere it can touch real systems, files or messages.
When I am describing a problem, asking a question, or thinking out loud rather than requesting a change, the deliverable is your assessment. Report your findings and stop. Do not apply a fix until I ask for one. Before taking any action that changes real state, check that the evidence actually supports that specific action.
It draws a clean line between advising and acting, which is exactly the line you want with a model this capable.
Download .mdHow do you stop it gold plating a simple job?
At full power Fable 5 can over-deliver: extra features, extra structure, extra polish nobody asked for. All of it costs credits.
When to use it: Add to focused fixes and small builds where you want exactly what you asked for and nothing else.
Don't add features, refactor, or introduce abstractions beyond what the task requires. A small fix doesn't need surrounding cleanup, and a one-off job doesn't need a reusable system behind it. Don't design for hypothetical future requirements: do the simplest thing that works well. Don't add handling for scenarios that cannot happen.
The simplest version ships faster, costs less of your weekly limit, and is easier to check.
Download .mdHow do you set up a run that finishes while you sleep?
Left alone, a long run can stall on a polite question like "Want me to continue?" and sit there for eight hours waiting for you.
When to use it: Use for any job you plan to walk away from: overnight work, weekend runs, anything unattended.
You are operating autonomously. I am not watching in real time and cannot answer questions mid-task, so asking "Want me to...?" or "Shall I...?" will block the work. For reversible actions that follow from the original request, proceed without asking. Before ending your turn, check your last paragraph. If it is a plan, an analysis, a question, or a promise about work you have not done, do that work now. End your turn only when the task is complete or you are blocked on input only I can provide.
It converts every would-be checkpoint into forward progress, so you wake up to finished work instead of a question.
Download .mdHow do you get Fable 5 to put helpers on the job?
Fable 5 can spin up subagents, which are helper copies that work on separate pieces at the same time. Most people never ask.
When to use it: Use in Claude Code when the job splits into independent pieces: research five competitors, audit ten pages, process a folder of files.
Delegate independent subtasks to subagents and keep working while they run. Intervene if a subagent goes off track or is missing relevant context.
Independent pieces get done in parallel instead of one after another, so the whole job lands sooner.
Download .mdHow do you make it remember lessons between sessions?
Every new session starts from zero. You end up re-teaching the same preferences and re-fixing the same mistakes.
When to use it: Set up once in Claude Code for any project you return to. Then run the second prompt whenever you finish a big piece of work.
Keep a notes folder for lessons learned. Store one lesson per file with a one-line summary at the top. Record corrections and confirmed approaches alike, including why they mattered. Don't save what the project files already record; update an existing note rather than creating a duplicate; delete notes that turn out to be wrong. To set it up from history, run this once: Reflect on the previous sessions we've had together. Use subagents to identify core themes and lessons, and store them in the notes folder. Make sure you know to reference it in future sessions.
Anthropic reports Fable 5 performs notably better when it can record lessons from previous runs and read them back later.
Download .mdHow do you get summaries you can actually read?
After hours of work the final summary can read like an engineer's shorthand: dense, jargon-heavy, full of labels you never saw.
When to use it: Add to any long job where the final write-up matters, especially work you will forward to someone else.
Lead with the outcome. Your first sentence should answer "what happened" or "what did you find", the thing I would ask for if I said "just give me the TLDR". Supporting detail comes after. When you write the final summary, write it for a reader who did not watch you work: complete sentences, terms spelled out, no shorthand or labels you invented along the way. If you have to choose between short and clear, choose clear.
The summary is the product. This makes it readable by the person who was not in the room, which is usually the person who matters.
Download .mdWhy should you never ask Fable 5 to show its thinking?
Old prompt packs are full of "explain your reasoning step by step" and "show your thinking". On Fable 5 these backfire.
When to use it: This is a warning, not a prompt. Check your saved prompts and delete these instructions before reusing them with Fable 5.
Avoid instructions like: "show your thinking", "explain your reasoning step by step", "transcribe your internal reasoning", "think out loud before answering". Fable 5 declines these and the request can fall back to a lesser model, so you pay Fable prices for Opus answers. If you want to follow along, ask for progress updates or a summary of the approach instead. The reasoning already happens on its own; you don't need to ask for it.
Fable 5 protects its internal reasoning by design. Asking for it triggers a refusal instead of a better answer.
Download .mdHow do you make your limits last the week?
Fable usage comes out of your plan's weekly limits, so the winners are the people who spend it deliberately. Four habits cover most of it.
One brief beats ten messages
Every message you send restarts its thinking. Put the goal, the context and the constraints in one proper brief and let it run.
Give it whole jobs, not fragments
Fable 5 is priced for depth. Sending it the same small asks you send a chatbot is the fastest way to waste a week's limit.
Keep everyday chat on Opus or Sonnet
The model picker exists for a reason. Routine questions get equally good answers from models that barely touch your limits.
Walk away and let it finish
Interrupting a long run to check in burns the work in progress. Set up the run with the prompts on this page, then leave it alone.
Want all 12 prompts in one file?
The full prompt pack, free
All 12 prompts in one markdown file. Drop it into your notes, your prompt library, or straight into Claude Code. No email, no catch.
Download the Full Prompt Pack (.md)
What else do operators ask about Fable 5?
Is Claude Fable 5 included in my Max plan?
Yes. Anthropic lists Claude Fable 5 as available to Pro, Max, Team and Enterprise users, with Fable usage drawing from a set share of your plan's weekly limits. The exact share has shifted during the rollout, so check the usage page in your Claude settings for the current numbers on your plan.
What is the difference between Claude Fable 5 and Claude Opus?
Opus answers your question. Fable 5 runs your workflow. It sits in a tier above Opus, plans before it executes, checks its own work, can spin up helper subagents, and keeps going for hours on a single job. For quick questions and everyday chat, Opus is still the right tool and costs far less of your limit.
Will Fable 5 burn through my weekly limits faster?
Per request, yes, because it works longer and more thoroughly than other models. The way to make limits last is to change what you send it: batch your instructions into one proper brief, give it whole jobs instead of a stream of small asks, and keep routine chat on Opus or Sonnet. One well-briefed long run costs less than ten badly-briefed short ones.
Do these prompts work in both the Claude app and Claude Code?
Yes. Every prompt on this page works as a paste-in at the start of a conversation in the Claude app or Claude Code. The subagent and memory prompts do their best work in Claude Code, where Fable 5 can create files and run helpers, but none of them require any technical setup.
Why does Fable 5 sometimes decline a request?
Fable 5 runs extra safety checks on a few sensitive domains, and it also declines any request to reveal its internal reasoning. If a normal business request gets declined, rephrase it in plain terms or run it on Opus. If you asked it to show its thinking, remove that instruction; see the last section of this guide.
Was Claude Fable 5 really pulled after launch?
Anthropic's product page shows Fable 5 was announced on 9 June 2026 and access was restored on 1 July 2026, so there was a real gap of several weeks. Anthropic has not published a detailed explanation, and the viral posts claiming to know the inside story are speculation. What matters for you: it is available now, inside plans you may already pay for.
Want to get properly good at this, in a room, on your own business?
The AI Operator Workshop is a hands-on day where you set up Claude, build your own workflows, and leave with systems running. Small group, real work, no theory decks.
See the AI Operator WorkshopOr see what we build with this every day: talk to one of our AI voice agents live.
Leo Garcia-Curtis
Founder & Head of Operations, Waboom AI
Runs Waboom AI on Claude every day, from voice agents to this website