Plans, limits and usage

Pangur has three plans you can be on: Free, Researcher and Institution. Each plan sets a few allowances, such as how many Ask questions you can ask and how much you can store. Separately from those, every plan has daily limits on model use, which exist to stop one account from running Pangur's models without end. This page lists both kinds of limit, says what each one counts, and explains the usage page where you can see how much of each you have used.

The figures here are Pangur's standard limits. An administrator can change a plan's limits, so the pricing page and your usage page show the figures in force on your deployment. Buying or changing a plan is covered in Billing.

The plansCopy Anchor Link

Prices include VAT. If you hold a seat in an organisation, you are on the Institution plan for as long as the seat lasts, whatever you have bought yourself, and if you have several sources of a plan, such as a subscription and a seat, the higher plan applies.

Plan allowancesCopy Anchor Link

What each allowance countsCopy Anchor Link

Ask questions counts the questions you ask in Ask, whether on the web, on your phone, in the Mac app or with pangur ask, over a rolling window. On the Free plan each question counts for 30 days from when you asked it, so a question asked on 3 May stops counting on 2 June. There is no single reset date. On Researcher and Institution each question counts for 24 hours. Two things do not count: answers that an assistant such as Claude writes and records in your Ask conversations, and pangur ask --materials, which returns passages without writing an answer.

Storage counts the files you upload or import into your library, your photographs, and the media in your exhibitions, across your personal projects. In an organisation's project, files count against the organisation's storage, which is 20 GB for each seat. Storage does not reset. The figure goes down when files are deleted for good, which happens 30 days after they go into the trash, or at once if you empty it.

Live exhibitions counts the exhibitions in your personal projects that are published at the time. Drafts and unpublished exhibitions do not count. Publishing is not offered on every account or deployment.

Recording transcription counts the minutes of audio and video that Pangur's speech model has transcribed for you since the first day of the calendar month (UTC), and it resets at the start of each month. The usage page lists it with the other allowances.

What happens at a limitCopy Anchor Link

When an action would pass an allowance, Pangur refuses it with one of these messages and changes nothing:

Daily model limitsCopy Anchor Link

Every plan has daily limits on the work Pangur's own models do for you. They exist to stop a runaway loop or a mistaken script from using models without end, but heavy work can reach them in ordinary use: analysing one 300-page volume uses about 220,000 input tokens, which is more than half of a day's model use. The limits apply on every plan and on every surface, which is to say the web, the apps and the command line.

A token is a piece of text, roughly three-quarters of an English word, that a model reads or writes. Model use covers Ask's answers, summaries and whole-work reads, the library map, reading more widely and summarising in Write, and exhibition composition, and analysis draws on it too. Long-form writing covers full write-ups, the longer answers that Ask's Expand produces in the background; a full write-up that is refused at the limit costs nothing. A manuscript reading sends one leaf's images to the vision model, and a transcription reads a leaf twice, so it counts as two readings' worth.

Some model work is charged to the owner of a project rather than to the person who prompted it, for instance the library map and a corpus's background reading. Which actions use whose allowance is set out in Models, assistants and who pays.

Limits for assistantsCopy Anchor Link

An assistant connected over MCP never runs Pangur's models, so neither the plan's Ask allowance nor the daily model limits apply to its work. The limits below do apply to assistants, as they do to everyone:

Uploads by an assistant also count against your plan's storage. The reasoning behind the exemption is explained under Who pays for an assistant's work.

Other limitsCopy Anchor Link

These limits apply on every plan.

The usage pageCopy Anchor Link

The usage page shows where you stand against each limit. Open Usage from the Account & billing page, from See usage in its Plan panel, from the list under Settings, or with the Go palette.

The page starts by naming your plan and, when the plan comes from an organisation seat, the organisation it comes from.

Plan limits has a row each for Ask questions, Model use today, Full write-ups today, Long-form writing today, Manuscript readings and Manuscript transcriptions. Each row shows how much you have used out of the limit and when it resets. For a limit that resets at a fixed time, such as midnight UTC, the row gives that time. For Ask questions it explains that each question counts for 30 days (24 hours on Researcher and Institution) and says when the oldest drops off. A row you have not used says so, and a limit your plan does not have is shown as unlimited.

Storage and exhibitions shows your storage and your live exhibitions. These are running totals, so they do not reset.

Last 30 days lists your use of Pangur's models by task, with the number of times and the tokens each used, and a total. It shows counts only, because Pangur does not show model costs in money. Assistant work does not appear here, since it uses no Pangur model.

If the page cannot read a daily count at that moment, it shows the count as unknown. While the daily counts cannot be read, Pangur lets requests through rather than refusing them.