Last updated September 18, 2026
Reliability
Your manuscript is the least replaceable thing you own. This page explains where it lives, what happens when something breaks, and what you can do about it without waiting for us.
Two copies of your work, always
As you write, PullAtlas keeps a complete copy of your project in a database on your own device, and syncs it to your account on our servers. The local copy is not a cache of the page you happen to have open. It is the whole project, and it is the copy the editor actually reads from and writes to. The copy on our servers is what carries your work to your other devices.
When your connection drops
Keep writing. The editor is reading from the copy on your machine, so losing your connection changes nothing about typing, opening a chapter, or moving around your Atlas. Your changes queue locally, and when the connection comes back they upload in the order you made them.
When PullAtlas has a bad day
If your workspace is already open, our services can fail without stopping you writing. Syncing pauses and the editor does not, which is the whole point of building it this way.
If you are opening PullAtlas fresh in a browser at the moment our site is unreachable, the page has to load from us before it can hand you the local copy, so you would have to wait for us to come back. That is the one case where being local-first does not save you, and we would rather say it than leave you to find it. A desktop app that lives on your machine and never has to fetch itself is the proper answer, and it is on the roadmap as a longer-term goal rather than a near one. Until it arrives, the protection that actually works is a backup, which is part of why they are free on every plan.
Nothing is dropped quietly
A change that cannot reach the server is held and retried, not discarded. If syncing cannot make progress, the workspace tells you so rather than showing you a tick it has not earned. We would much rather you see an honest warning than find out later that a session went nowhere.
Your own copy, free on every plan
You can take a full backup of an entire project, prose and all, at any time, on any plan including Free. You can also export a clean manuscript file to read, submit, or hand to someone. Neither is a paid feature, because your words leaving with you is not something we think you should have to buy.
If PullAtlas ever went away
We are a young company, and the honest thing to do with a young tool is tell you what your exit looks like before you need it. It is not a favour we grant on the way out: backups and exports are free and always available, the files they produce are ordinary formats other software reads, and deleting your account is one click in Settings rather than an email you have to chase. Nothing about your work is held hostage to us continuing to exist.
Live status
The current state of PullAtlas is always at status.pullatlas.com. The website, the workspace, sign-in and sync are each checked automatically every minute from three regions, and the page updates itself when one of them stops responding, so a problem shows there without anyone having to notice it first. It also shows the status of Paddle, Anthropic and OpenAI, taken from their own status pages.
A check can only see what it measures: it knows whether a service answers, not whether every feature behind it is working. If something is wrong for you while the status page is green, tell us. That is exactly the kind of problem we most want to hear about.
What you can count on
Everything on this page: a complete copy of your work on your own device, changes that are held and retried rather than dropped, backups and exports free on every plan, and a status page that reports on its own. What we do not offer yet is a formal uptime guarantee or service level agreement on any plan, and we will not publish an uptime figure until the status page has a record long enough to stand behind one.
Telling us something is wrong
Email support@pullatlas.com and you will reach the person who wrote the code. If you can, say what you were doing and roughly when, which is usually enough to find it. For what we store and who processes it, see the privacy policy.