Show your sources, ask before you act: how Plexii is designed
An AI assistant is only useful at work if you can trust what it says and stay in control of what it does. Plexii is built around two rules that follow from that.
Rule one: show the sources
Plexii answers from your own documents, sheets, notes, tasks and tables. Every claim carries a citation you can click through to the item it came from, and a live trace shows what it read while it works. When it can't find something, it says so rather than inventing an answer.
That changes how you use it. You don't have to take the answer on faith; you check the two or three sources it points to, which is far quicker than doing the research yourself.
Rule two: ask before acting
When an answer implies a next step, Plexii drafts it as a card: a task, a table, a calendar block, an updated slide, a whole document. Each card shows a preview of the real change. Nothing lands until you apply it, one card at a time or all at once.
The busywork disappears. The decisions stay yours.
For bigger jobs on Pro and Team, Plexii can do multi-step work for you: browsing pages, gathering data into a table, drafting a set of changes. That work runs behind a consent gate, with a live run you can watch, a running cost and a Stop button that always works. Before it touches a new site it asks whether to allow it once, always, or not at all. It never signs in, pays or moves files on your behalf.
Your key or ours
You choose how the assistant runs. With your own Anthropic key, prompts go straight from your machine to Anthropic with no proxy. With managed Plexii on Pro, requests run through our server, which records token usage for billing and never uses your prompts for training. Either way, usage and estimated spend are visible in Settings.