Live demo
Drive it yourself.
The real desktop, the real apps, filled with demo data. Open Mail and read a thread. Watch an agent wait for your approval. Press ⌘K to search.
Nothing here is connected to a real account — it's a sandbox you can't break.
search_toolsfound gmail.messages.listget_tool_detailsread the schemaexecute_toolscanned 1,284 messagesawaiting approvalunsubscribe from 23 sendersRe: Q3 board deck — a few notes
Priya Raman · to me · 8:47
AI summary
Priya wants the retention chart moved before pricing, flags slide 14 as too dense, and needs the final deck by Thursday 5pm.
Hi — went through the draft last night. Three things:
1. Retention is our strongest number this quarter. It should lead, not sit after pricing.
2. Slide 14 is doing too much. Split the cohort table out.
3. I need the final by Thursday 5pm to circulate before the meeting.
Thanks — P.
Like what you see? The real thing connects to your actual tools.
Curious why it's this open? Read why we let you break the demo and how it was invisible to Google for weeks.
What's inside the demo
Eleven apps, all of them AI-native and all of them operable by your agents. Open any of them from the dock above — each one is filled with realistic data so you can see how it behaves in use, not just how it looks.
Chat
One thread across every model. Ask in plain language; it picks the tools, shows its plan, and waits for approval before anything irreversible.
Every thread arrives summarised, with deadlines and asks pulled to the top. Drafts replies from the conversation, not from a template.
Calendar
Your agent defends focus blocks, prepares agendas from recent threads, and explains every change it makes to your day.
Files
Documents, recordings and folders your agents can read. Calls get transcribed and summarised on arrival, so search finds what was said.
Docs
A writing surface with an editor that suggests structural edits — move this section, split that table — and applies them on approval.
Sheets
Spreadsheets where a cell can be an =AGENT() call. Ask for churn per plan and it reads the source material and works it out.
Agents
Every run is listed, replayable and scoped. Open one to see each tool call it made, in order, with what it returned.
Workflows
Install a prebuilt automation in a click — inbox triage, meeting prep, invoice chasing — or build your own from the same pieces.
Connectors
A thousand tools as searchable data, not hardcoded plugins. Agents find one by meaning, read its schema, then run it through a fail-closed executor.
Studio
Generate and iterate images from a prompt, with variants side by side and upscaling built into the same window.
Settings
Model routing, approval policy, data residency and the credential vault — including the local filter that keeps trivial prompts off the network.
Questions about the demo
Is this the real Velaris interface?
Yes. The demo runs the same desktop, the same apps and the same interactions as the product — the only difference is that the data is fictional and nothing connects to a live account.
Do I need an account to try it?
No. The demo runs entirely in your browser. There is no sign-up, no install and no backend call, so nothing you do here is stored or sent anywhere.
Can I break anything?
No. It is a sandbox with demo data. Close windows, approve agent runs, install workflows — reloading the page resets everything.
What is an AI operating system?
A single environment where your apps, your autonomous agents and the tools they use all live together, instead of an AI chatbot bolted onto separate products. Agents can operate the apps directly, with your approval on anything destructive.
Does it work on a phone?
Yes. On touch devices the demo switches to a stacked, one-app-at-a-time layout, because dragging windows on a small screen is worse than not having them.