Our story
DunOps was built by developers who kept shipping products faster than they could manage the infrastructure behind them. This is the story of what we built about it — and why.
We were shipping apps in hours but losing days to DNS panels, deploy dashboards, and infra configs that had nothing to do with the product we were building. We wanted to talk to DevOps. So we built something that could listen.
— DunOps founding team
By the numbers
1weekend
to build the first working prototype
100%
of mutating actions require your approval
0passwords
stored — ever. Auth is codes and links only
∞
times you can replay a saved Playbook from chat
How we got here
Apps were flying out the door with AI. But every deploy turned into a four-tab archaeology session — Vercel dashboards, Namecheap DNS panels, GitHub Actions logs, and a Slack thread where someone had the password from 2022. We were spending more time managing infra than building product.
"What if DevOps was just a chat thread?" We built a prototype over two days. It could deploy a Vercel project, point a domain to it, and report back when DNS had propagated — all in a single conversation. We showed it to five friends. All five said the same thing: I need this now.
An AI agent with direct infra access is terrifying unless you trust every action it takes. We didn't ship a single feature before we shipped the approval gate. Every plan is shown, every diff is surfaced, every mutation waits for you to say yes. That constraint became the architecture.
Users kept running the same three or four sequences — deploy to preview, connect domain, promote to production, recover a rollback. We built Playbooks: a visual canvas where you author a flow once, save it, and replay it from any chat thread forever. The feedback loop collapsed from 30 minutes to one message.
DunOps shipped to the world. The core promise is unchanged from that first weekend prototype: if you can describe it in plain English, DunOps should be able to do it — after you say yes.
What we believe
These aren't marketing copy. They're the constraints we built the product around — written before any line of product code.
No operation that changes infrastructure runs without your explicit approval. Not one. This is not a setting you can disable — it is the architecture.
You see every plan, every diff, every API call before it executes. DunOps doesn't hide complexity — it organises it so you can make informed decisions in seconds rather than minutes.
Approving one step does not authorise the next. Each consequential action is a separate decision point. We believe that's how powerful tools should work.
We build for people who ship ideas fast with AI. You shouldn't need a DevOps background to run production infrastructure. We're living proof that you don't.
Where we are
DunOps is in public beta. The surface area keeps growing — more providers, more workflow types, richer playbook tooling — but the foundation is fixed: your infrastructure, your approval, your timeline.
We're not here to automate you out of the loop. We're here to make the loop faster. If you build with AI and your deploys still feel like archaeology, this was built for you.
What's next