Product screens recreated with synthetic data.
nventr Agent is that definition on a page you already have. Agent IO gives you three installs on one token: a script that floats a launcher, an iframe that fills a fixed panel, or the session API if you draw the chat yourself. It speaks in the role you wrote, in the accent and on the side you chose. It can read the records you allowed. When it wants to create something — a project, a pipeline, a workload — it shows a proposal. The record appears after a person confirms, and not before.
That confirm step is the whole safety story for this surface. The model is not trusted to create. The button is the commit. Rotate the token in Agent IO and this copy of the widget stops being valid.
It is one agent in one conversation. It does not split itself into specialists, and it does not post an invoice to an ERP. Several specialists on one objective are an nQube coordinator block, defined from this catalog, and a commit still waits on a gate. Use this embed when the job is a person on your site who needs an assistant with a confirm button.
nventr Agent is not agentOS, and it is not Agent IO. Agent IO is the definition screen, including the specialist library. agentOS commits workflow writes. This is one conversation on the page. Multi-role work is the nQube coordinator block, not this embed.
A team that wants an assistant on a site they already have — an intranet, a portal, a product page — without giving that assistant a write credential.
Platform, as the embed for an agent you defined in Agent IO. The page it sits on can be yours. You do not need an App Kit instance to embed it.
Pricing →Synthetic example. Not a customer.
The help page loads the script. Someone asks, “Set up a pipeline for vendor invoices.” The agent answers with what it would create — name, workspace, a one-line description — and a confirm button. They confirm. The pipeline record exists, empty, ready to design in nQube. They ask it to post invoice 20841 to the ERP. It cannot. That action was never a tool on this agent. The ERP write, if you want it, is a workflow.
A script tag, an iframe, or three API calls. No new app to rent for the widget itself.
Creates wait on a button. A chat reply is not a write.
Change the prompt or the toggles in Agent IO and this embed follows. Rotate the token and it stops.
Agent IO holds the prompt, the reads, and the proposals. Without that, there is no agent to embed.
Agent IO gives you a script for the floating launcher, an iframe for a fixed panel, and the session API. All three use the same token.
It answers in the role you wrote, using the reads you allowed.
A create stays a proposal until a person commits it. Closing the widget does not sneak the write through.
Platform Start is the self-serve canvas. A first pipeline we stand up with you is delivery, priced apart from the subscription.