Frequently asked questions

It can be told, and it can fetch, but neither is remembering. Paste a report and the AI holds a snapshot: stale the day your numbers change, with no way to verify it. Connect a raw data connector and it fetches live, then forgets everything when the session ends and re-derives it all differently next time. What doesn’t exist is the middle thing: a maintained, verified memory of your business, computed once, checked against your store to the penny, kept current, and identical every time anyone asks. That’s Keeper.

One OAuth click per source, then about two hours while Keeper compiles two years of memory and reconciles it against your store. You can watch it check its work, and you’ll be notified the moment your memory is live.

Your data lives in Keeper’s infrastructure, encrypted and scoped to your workspace: exactly one brand’s team can see one brand’s memory. Claude and ChatGPT never receive or store your customer database. They send questions, Keeper sends back answers: counts, metrics, findings, never customer records. Nothing is ever used to train any AI model. Audience syncs travel directly from Keeper to Meta or Klaviyo, server to server, never through the chat. And if you disconnect, we delete the compiled memory. On free, everything is read-only; nothing writes to any system until you enable actions on a paid plan.

No. The memory lives in Keeper, not in the AI. When anyone on your team asks a question, the AI sends the question to Keeper and receives an answer: a metric, a finding, a count, never a customer list. When an audience syncs to Meta or Klaviyo, it travels directly from Keeper to that platform. The chat only ever carries questions and answers, which is exactly what you can tell a security reviewer.

Keep them. Those are dashboards with an AI door. Keeper is memory for the AI you already talk to every day: compiled, versioned, reconciled, and served inside Claude and ChatGPT. It’s free to connect and verify against what you have.

Yes. Keeper’s commercial intelligence is free. Connect your data and Keeper builds and maintains your commercial memory, refreshes it daily, and lets you explore your findings and personas. Paid plans unlock actions — such as syncing audiences to Meta and drafting campaigns or flows in Klaviyo.

Keeper refreshes your commercial memory daily on both free and paid plans, so your AI is working from current business context rather than a one-time analysis.

You pay when you want Keeper to act. Paid plans unlock execution inside connected tools, including Meta audience sync, suppression audiences, and Klaviyo campaign and flow drafts.

Sources: Shopify, Klaviyo, Meta, reviews, and loyalty. Reading side: Claude and ChatGPT, through their native connector support. Actions write to Klaviyo and Meta. If a source you rely on is missing, tell us on a walkthrough call; the roadmap is shaped by exactly that.

Keeper is built Shopify-first today, and the to-the-penny reconciliation is anchored to your Shopify store. If you’re on another platform, book a walkthrough and tell us; other platforms are on the roadmap and demand decides the order.

No. On free, everything is read-only. On paid, actions run only when you ask for them in the conversation, and built flows land as drafts your team reviews before anything ships. Keeper never acts on its own schedule.

Those connectors give your AI raw access to one platform at a time, and it re-derives everything from scratch each session. Stack one per client and you get twelve pipes, not one memory: every cross-brand question re-fetches every account, hits rate limits at scale, and sees only that platform’s slice. Keeper compiles all your sources into one reconciled, versioned memory, so answers are instant, cross-source, and identical every time you ask. Access is the pipe. Memory is the point.

They’ll match your Shopify store to the penny; that’s the contract. They won’t always match Klaviyo’s or Meta’s dashboards, because each platform attributes revenue its own way, and that disagreement is exactly the problem Keeper resolves. Every number comes with its definition and compile date, so you always know what you’re looking at.

Personas need enough customers to cluster with confidence. Larger bases generate cleanly; smaller ones generate with confidence indicators so you know how much weight to put on them; and if your base is genuinely too small, Keeper says so instead of inventing segments. Honest limits are part of the product.

Yes. Every brand you connect gets its own memory in one workspace, and seats are never metered. If you’re running brands for clients, the partner program is built exactly for that: one workspace, your whole book.

Your data stays yours, always. Cancel a paid plan and the brand drops back to free: memory keeps refreshing daily and actions turn off. Disconnect entirely and Keeper deletes the compiled memory for that brand.