Online models and plans
Online models are optional. A paid plan adds frontier models from multiple providers and web search with cited sources - the local app never depends on any of it, and canceling changes nothing local.
How it works
Online models need your Flowsta identity, which lives in the Flowsta Vault on your device: free, about two minutes to set up, 24 recovery words to keep on paper, and no password for the Vault itself - nobody can lock you out. Settings, the Online Models page and the welcome's last step tell the same story before you click. Get Flowsta Vault - free opens a handoff page that brings you back to Your Own AI the moment the Vault is ready; when the Vault is installed but closed, the same places offer Open Flowsta Vault and notice the moment it is ready; a locked Vault is told to unlock first, and a sign-in waits for it instead of falling back to a separate sign-in in the browser.
Signed in, See plans opens yourownai.net already signed in as you, with this computer linked to your plan, and the dashboard says who is signed in. The first time, Flowsta's "Share with Your Own AI?" page asks once; Allow carries on to the dashboard. If you disconnect Your Own AI in your Vault, Settings says so instead of quietly signing out.
Then choose a plan, and the Online Models page lists the catalog: filter by chat, web search, or coding; every card shows its context window and prices per million tokens up front, plus any per-request or per-search fees, and the model's catalog name is the one the model menu uses. Each card has an Auto switch for whether automatic routing may use it, a Use for menu that makes it your pick for a kind of question, and badges saying what routing uses it for - "your pick" or "recommended". A model that is failing for many people right now says "Having trouble right now" and is left out of automatic picks until it recovers; a model you picked by name is still used. Pause any model to keep it out of routing and out of the model menus; an AI already set to it keeps it.
Which model takes which job
An Online and Offline AI sends each question to the online model for its job - Everyday questions (GPT-6 Luna by default), Hard questions, Needs current information, and for projects Agent work and Planning. Each has a recommended model you can change in Routing settings. The recommendations follow the online catalog: a model the catalog marks as the default for a job takes that slot, so new models and providers arrive without an app update.
When a provider is down
When an online model's provider is busy, erroring or not answering before its first word, an Online and Offline AI asks a model of the same kind from another provider you allow, once, and the reply's model note says which and why: "online: GPT-6 Luna was busy, so DeepSeek V4.1 Flash answered". An AI set to one online model is never switched: its error offers Ask <the other model> instead beside Retry. Settings › Routing lists the providers automatic routing may use, each with where it is based, and shows Where your answers came from: your recent questions by where they were answered, and this plan period's online cost by kind of model. See Routing settings.
Privacy, plainly
Your message is forwarded through Flowsta's relay to the provider with your identity stripped - the provider receives the content but not who you are. The relay passes messages through without storing or logging them; only usage amounts (model, tokens, cost) are recorded for billing. Attachments stay on your device unless you turn on one setting - see Attachments and images - and health questions go online only as an asked-every-time choice (Health questions).
Allowance and metering
Each plan includes a usage allowance, at least what you pay: $20 each month on Pro, $63 on Power (your $60 plan plus $3), $220 on Ultra (your $200 plan plus $20). The full amount is there from the day you join and resets on that date each month. An annual plan is one month free, with the year's usage spread evenly over its twelve months ($18.33, $55 or $183.33 a month). Usage beyond the allowance is metered at the rates shown in the app, with an overage opt-in you control from your Flowsta identity page. A plan bought before 1 October 2026 keeps the allowance and billing date it was sold with until it renews.
When the allowance runs out mid-conversation, the app shows a plain notice with your options - enable overage, upgrade, or wait for the reset. Everything local keeps working regardless.
How overage is billed
Past your allowance you pay the same rates. Your card is charged whenever accrued overage reaches $100 ($500 on Ultra), and again when the period ends. If a charge fails, online models pause past your allowance until the invoice is paid - the Flowsta identity section in Settings says so, names the period and the amount, and offers Pay the invoice. Use within your allowance, and everything local, keeps working.