Agent-native Architectures: How to Build Apps After the End of Code

Was this newsletter forwarded to you? Sign up to get it in your inbox. Plus: Help us scale the only subscription you need to stay at the edge of AI. Explore open roles at Every.

Any application you use daily—whether it’s Word, Figma, or Gmail—is a bronze and glass facade towering 500 feet above the street. It is a lobby with travertine walls that smells faintly of sandalwood. Every beam is load-tested. Every force and flow obeys the blueprint.

Just to be real with you, I am jealous of architects. I often moonlight as one, but as a programmer, my skyscrapers are shoddy. I start before the blueprint is final; I dig a foundation and sink some beams, but they are usually off by an eighth of an inch. By the time we get to the fifth story, I need a real architect to take over.

But AI enables a new kind of software, one that’s more like growing a garden than it is building a skyscraper. I’ve been calling it an agent-native architecture—and we’ve pivoted our whole software strategy at Every around it.

The core of an agent-native architecture is not code. Instead, as the name implies, the core is an agent—something squishy and alive, planted in sun and soil. Each feature of the app is a prompt to the agent that names the result to achieve, not a set of steps to follow. That’s why I often think of agent-native apps as Claude Code in a trenchcoat.

Because the agent handles the how, developers only have to name the what. This makes apps faster to build, fix, and change. It also makes them malleable: Users can alter how the app behaves just by changing words in a language we already speak. This levels the playing field of power. Software becomes something we build together, not something only a rarefied few can do.

A gardener clips and weeds, but ultimately gardens grow into something that cannot be anticipated or specified. They are wild, flexible, and free. And for someone like me—and for millions of others—it’s the first time that the way I build isn’t a liability.

Agent-native architectures are terrifying and unintuitive from an architect’s perspective.

“How can we allow software so much freedom?” This, of course, is really a question about users and ourselves.

I wonder if, in a few years, living in a garden of agents, we’ll look back at our era of tall, precise, and perfect skyscrapers wistfully. A gorgeous, elaborate attempt to impose total control on a world that doesn’t want to be controlled.

The complete guide to agent-native architectures

Today, we’re publishing a complete guide to agent-native architectures on Every.

I wrote it with Claude, and it has everything from a high-level breakdown of agent-native architecture principles to low-level implementation details. If you—or your agent—want to become an expert on agent-native architectures, you should read it:

We’ve also published this guide as a skill in our compound engineering plugin for Claude Code.

Dan Shipper is the cofounder and CEO of Every, where he writes the Chain of Thought column and hosts the podcast AI & I. You can follow him on X at @danshipper and on LinkedIn, and Every on X at @every and on LinkedIn.

To read more essays like this, subscribe to Every, and follow us on X at @every and on LinkedIn.

We build AI tools for readers like you. Write brilliantly with Spiral. Organize files automatically with Sparkle. Deliver yourself from email with Cora. Dictate effortlessly with Monologue.

We also do AI training, adoption, and innovation for companies. Work with us to bring AI into your organization.

Get paid for sharing Every with your friends. Join our referral program.

For sponsorship opportunities, reach out to [email protected].