September 23, 2026 / 4 min read
Duvi Is Finally Live
A year ago Duvi was a chatbot I thought would take a week. Now it is a product anyone can use, and launching it on Product Hunt made the whole strange journey feel real.
- duvi
- launch
- product hunt
- ai agents
- voice
Duvi is finally live.
That sentence feels much smaller than the year behind it.
A few weeks ago I wrote about how Duvi started as a chatbot I gave myself one week to build. The week became infrastructure, voice, WhatsApp, phone calls, tools, retrieval, latency budgets, and a long list of things I only learned because the earlier version broke in front of real people.
At the end of that post, Duvi Studio was still opening soon.
It is open now.
The best part is how little someone has to know
You can go into Duvi, describe the agent you want, and give it a URL.
That is it. Duvi reads the site, builds the knowledge, writes the first version of the prompt, and gives you somewhere to test it before a customer ever sees it. If it is a store, the agent can connect to the catalogue and check what is actually in stock. If a question needs a real answer from another system, it can call that system. If the customer wants something done on the page, the agent can do that too.
Then the same agent can answer on your website, your phone number, and WhatsApp.
I have spent so long thinking about everything underneath that it is still slightly strange to watch the product make those decisions disappear. A person does not need to know how retrieval works. They do not need to choose a chunk size, plan a retry policy, or understand why a voice pipeline feels slow. They should be able to say what the agent needs to do and get on with the part of the business that only they can do.
That is the product now.
Product Hunt made it public in a different way
We launched Duvi on Product Hunt with a simple line: make an AI for any business just using a URL.
I have launched things before, including Vox earlier this year, but this one felt different. Duvi already had a history before it had a public launch. The underlying work had answered real customers for real brands. I knew it could survive traffic. What I did not know was whether a stranger, arriving with none of that context, would understand why it should exist.
People did.
They tried it, asked questions, followed the launch, and gave us the kind of feedback you cannot manufacture inside your own team. The launch went well, but the number on the page is not the part I keep thinking about. The useful part was seeing people understand the idea quickly: one agent that knows the business, connects to the tools it already runs, and works everywhere customers want to talk.
For a product built out of a year of complexity, being understood in one sentence felt like the win.
The idea got bigger while the product got simpler
The first version answered questions.
Then it needed to speak. Then it needed to work on WhatsApp and phone calls. Then it needed live information instead of a frozen copy of a website. Then it needed to take an action, save what happened, and let a team ask about the result later.
Every addition could have turned Duvi into a collection of separate products. A website chatbot over here, a voice agent over there, an automation builder somewhere else.
The work was making all of it feel like one thing.
One agent. One set of knowledge. One set of tools. Every way in.
That is also why the conversational builder matters so much to me. The product became more capable, but the setup became a conversation. The more we built underneath, the less we wanted the user to see.
Live is a beginning, not a finish line
There is an odd temptation when you launch something to describe it as finished. It makes the announcement cleaner.
Duvi is not finished. It is live.
Those are different things.
Live means someone can create an account today, build an agent, point it at their own content, test it, and put it in front of customers. It also means the questions are no longer hypothetical. People will connect systems I did not expect, ask for things I have not designed yet, and find edges that a year of testing could not reveal.
I am excited about that part.
For most of the last year, Duvi was a thing we were building because we had seen the problem ourselves. The Product Hunt launch was the first time it properly belonged to people outside that story.
The one-week project is live. Now we get to find out what everyone builds with it.