A run stopped before it finished and shows "Keep going": your Claude plan's usage window ran out

You started Otto this morning and got this: a run that stopped on its own partway through, with a Keep going button sitting next to the status line. The wording reads close to "The Flywheel paused at your Claude plan's usage limit - run it again after the limit resets and it picks up from its log." The reactions are always the same three, in this order: it paused before it finished, it keeps asking me to keep going, and is it eating my Claude plan. On the version shipping today the answer is one sentence long. That button means your own Claude plan's usage window ran out. Booked Solid did not stop the run.

What the Keep going button means today

Booked Solid shows Keep going in exactly one situation now: a run that stopped because your Claude subscription had no usage left in its current window. Nothing else in the app interrupts a run partway through. There is no Booked Solid spending limit, no token ceiling of ours, and nothing that halts a job because of what it cost. If a run stopped and offered you that button, your plan hit its own wall.

Alongside the button you will usually see a line from the run itself, worded close to "Your Claude plan's usage window is used up for now - that is the plan refilling, not a fault in the run. Nothing is lost: it keeps its log, so run it again after the limit resets and it picks up where it left off." That is the whole diagnosis. It is not a crash, it is not a bug, and it is not something a support ticket can unlock. Claude limits are Anthropic's, on your account, and only time or a bigger plan moves them.

What to do about it

1. Wait for the window to reset. Claude's own message is the thing that names the reset time, so read it and go by that. We deliberately never guess a reset time for you, because guessing one wrong is worse than saying nothing.
2. Then run it again. If the app is still open on that run, tap Keep going and it carries on the original ask from where it stopped. If you closed the app in between, just start the same job again. Either way it reads its own log first, so it picks up rather than starting from scratch.
3. Look at what is already banked before you re-run anything. Otto saves as it works, and The Flywheel writes the thank-you first, so usable drafts are often already sitting there. Check that gig's own folder under leads/, for example leads/your-gig-name/, and check your Board and your Inbox. If what you wanted is already written, leave the run stopped.
4. If it happens most days, the plan is the problem. Claude Pro at about 20 dollars a month covers most performers. Hard daily Autopilot use is what pushes people onto Claude Max at about 100 dollars a month. Moving up a tier is the only real fix for a plan that runs out every afternoon.

What actually spends your Claude plan

Everything a turn does spends it: reading your files, reviewing drafts, tool calls, and retries. That is the part people find surprising. Asking Otto to re-check work it already did costs roughly what fresh work costs, because it has to read everything again to answer you.

To see where a day actually went, open Advanced on the right of the top bar and choose Token history. It shows what each run used over time, names the tasks, and will export a shareable report. That view is usually enough to spot the one job that ate the afternoon.

How to make a plan go further

• Ask for one thing, not a sweep. If you already know about one specific reply, running a full scan inbox to deal with it is the most expensive possible route. Name it instead: "Handle [person]'s reply on the [thread] only. Draft the response and log it, do not scan the rest of the inbox."
• Point a Play at one gig by name. "draft the thank-you and review ask for the [gig name] gig" stays small. The Flywheel launched broadly on a busy workspace has to work out for itself which finished gig you meant, and that hunting is what costs.
• Start a fresh chat for a new topic. A long conversation re-reads its whole history on every message. Otto will tell you when a tab has got heavy: "This conversation has grown long, so each message now re-reads a lot of history on your Claude plan." When that appears and the new question is unrelated, open a new chat rather than carrying on.
• Run light on purpose if you want to. Under Settings > Plays that may start themselves, leaving those toggles and scheduled tasks off and running Plays by hand is a legitimate way to use Booked Solid. It spends a fraction of a plan, and a 20 dollar Claude Pro plan goes a long way that way.

Historical: the old "Paused to protect your Claude plan" card, versions 1.11.16 to 1.11.41

This section is history. It does not describe the app you are running unless you are on an old build. It is here because plenty of people remember it, and because the wording still turns up in searches.

Between 2026-07-24 and 2026-08-04, Booked Solid had its own spending brake. On a focused job it gave a heads-up at around 100k weighted tokens and then genuinely stopped at around 200k with a card reading "Paused to protect your Claude plan" and a Keep going button. A full sweep got much larger lines, roughly 500k and 1M. From 1.11.34 there was also a control at Settings > General > When to stop and ask with three choices: Recommended, No limit and Custom.

All of it was removed in 1.11.42. The card is gone, the two lines are gone, and the When to stop and ask setting is gone from the app entirely. It was removed because the line could not be drawn anywhere honest. 200k was chosen as the allowance for a narrow typed task, on the reasoning that a one-liner costing that much has gone wrong, but the same line applied to a whole Play. Three complete Silence Patrol runs on one desk measured 100k, 205k and 204k. An ordinary successful run was sitting right on the line and stopping about half the time. A brake that fires at the median cost of a working feature is not protection, it is a fault you have to keep clearing.

Three questions we were asked repeatedly at the time, answered here so the old threads make sense:

• "Noticed token limit before stopping was 200k yesterday and now it's 100k?" and did you lower my token limit: nothing was ever lowered. Those were two different messages, not one moving limit. 200k was the pause, 100k was the earlier heads-up that did not stop anything.
• Why it read 1618k and then 2030k: every tap of Keep going moved both lines up past what had already been spent, so the displayed figures were a running total for the whole task, every resume included. Nothing was drifting or broken.
• "Token issue doing Autopilot and then doing flywheel again": a Play started straight after an Autopilot sweep in the same session could stop at the narrow line while showing the sweep's continuation wording. That mismatch cannot happen now, because neither line exists.

Two things people wrote in about back then still hold, for a different reason. "Got this while Otto was reading a new reply and creating a draft from it" and the draft was already filed: that is still the good outcome, and it is still worth checking your drafts before spending more. And a Flywheel out of control, repeating itself and finishing nothing, was never really a token problem. It is a run failing to converge, and it is covered in The flywheel or lead engine keeps looping: how to stop it and stop the token drain.

Still seeing the "Paused to protect your Claude plan" card?

Then you are on an old build, and you should update. Check Settings > Updates, which tells you which version you are running. Anything from 1.11.42 onward has no spend brake at all, so if the number shown there is lower than that, take the update.

If it still does not work

Be blunt with us about which one you are looking at, because the two are fixed in opposite directions. A run that stopped at your Claude plan's limit needs time or a bigger plan, and there is nothing we can do to speed that up. A run that stopped for any other reason is ours to look at.

Generate a support bundle (see Generate a support bundle and turn on support logging) and open a ticket at support.booked.kivimedia.co. Include a screenshot of the stop message, the task name at the top of the run, and the Token history entry for that run. Those three together are usually enough for us to answer in one reply.

Still stuck? Email bookedsolid@kivimedia.freshdesk.com and a person will help.