4 Power Prompts to Try With NOVA

Most people ask their AI to write a paragraph and call it a day.

That is not what NOVA (Native Omnimodal Voice Agent) is for.

NOVA is a local-first, desktop-native voice agent you own outright. It runs on your machine, drives your apps, searches the live web, builds automations, and chases a multi-step goal across many steps without you babysitting it.

Below are four NOVA power prompts that show what that actually looks like. Each one is copy-paste ready. Say it out loud or paste it in. Watch the difference.

1. Computer use: make NOVA drive your desktop

This is the one that makes people sit up.

NOVA does not just talk about your computer. It runs it.

On Windows it has real UI Automation control. It finds named controls in the actual window tree, clicks them, types into fields, presses keys, and verifies the result instead of guessing at pixels.

"Open the Settings app, go to the Display section, and turn on Night Light. Then take a screenshot of the result so I can confirm it worked."

Why this is powerful: NOVA uses its UIA tools to locate the control by name, click it, change the value, and screenshot the foreground to prove the change landed.

No vague "here is how you would do that." It does it, then shows its work.

Once you trust that loop, you stop doing the small clicky chores yourself. You describe the outcome. NOVA performs it.

For more of these, see 12 Cool Things You Can Do With NOVA.

2. Web research: real answers from the live web

NOVA runs its own private web search through a local SearXNG sidecar.

No black-box API deciding what you get to see.

In deep mode it fetches the top results, extracts the readable content, reranks it, and can hand a research-shaped question to a sub-agent that synthesizes an answer.

That means you can point it at something current and get a grounded answer with the sources attached.

"Research the three best mechanical keyboards under $150 released in the last year. Pull current prices and reviews from the live web, compare switch type and build quality, and give me a short ranked list with a link for each."

Why this is powerful: this is not a model guessing from training data that went stale months ago.

NOVA fetches, extracts, and reranks live pages, then gives you a ranked answer you can actually click through and verify. You get the conclusion and the receipts.

Want the full picture on how the search stack works? Read NOVA Can Search the Web.

3. Automation: build an n8n workflow by voice

NOVA ships with n8n running as a local Docker sidecar and a set of tools to drive it.

It can list your workflows, read them, create new ones, update them, and toggle them on or off. You describe the automation in plain language; NOVA builds the workflow for you.

"Create an n8n workflow that runs every morning at 7am, checks my RSS feeds for new posts, and sends me a summary. Then turn it on and show it to me in the Automations tab."

Why this is powerful: automation used to mean dragging nodes around a canvas and reading docs for an hour.

Here you say the outcome and NOVA wires it up.

The workflow lives on your machine, in your n8n instance, under your control. You can open the Automations tab and see exactly what it built.

This is where a voice agent stops being a toy and starts buying back your time. Go deeper in How to Make Automations With NOVA and n8n.

4. A multi-step mission: set the goal, let it run

The first three prompts are single moves. This one is a campaign.

NOVA has a real mission system. You set a mission, break it into objectives, mark one active, and it works toward completion across many steps.

It can chain web research, computer use, and file writing into one coherent push instead of needing a fresh instruction every thirty seconds.

"Set a mission: plan a weekend trip to Portland. Objective one: research three neighborhoods to stay in and pick the best for walkability. Objective two: find two coffee shops and two dinner spots near that neighborhood. Objective three: write the whole plan to a file called portland-trip.md. Work through the objectives in order and tell me when each one is done."

Why this is powerful: you are not feeding it one prompt at a time.

You hand it a goal with structure, and it drives the whole thing; researching the web, making decisions, writing the result to disk, and reporting progress as it clears each objective.

That is the difference between a chatbot and an agent.

This is the closest thing to the assistant you imagined when you first heard the word. See How NOVA Is a Lot Like JARVIS From Iron Man.

The thread running through all four

Look at what these prompts have in common.

None of them ask NOVA to describe a task. They ask it to do the task.

Drive the app. Fetch the page. Build the workflow. Run the mission to completion.

That is the whole point of a desktop-native agent. It lives on your machine, with bring-your-own-key access to the models you choose, and it acts on your behalf in the real software you already use.

One time you pay for it; you own it. No subscription clock ticking.

Start with prompt one. Get the feel of NOVA actually doing something. Then climb to the mission and watch it run on its own.

If you are still setting up, the beginner setup guide walks you through it. And when you are ready to own one, Get NOVA.

The prompts above are a starting line, not a ceiling. The agent does more than you will think to ask for. Go find the edges.