PROGRAM

The creator of Vue opens the conference with his annual look at where Vue stands and where it’s heading. Evan covers the latest in Vue core, including Vapor mode and the next steps for performance. He also covers the evolving Vite toolchain that powers it, and what’s ahead for the Vue ecosystem in the year to come. A must-see for anyone building with Vue today, and for anyone planning what to build with it next.

Daniel Roe, lead of the Nuxt team, shares the state of the Nuxt ecosystem in 2027. He covers the road to Nuxt 5, what the new Nitro server engine means for your apps, and the latest features, modules and tooling from the past year. He also gives an honest look at where Nuxt is going next. Whether you’re upgrading an existing app or starting fresh, this is your guide to building full-stack Vue in 2027.

AI agents no longer just autocomplete code. They scaffold, run, debug and deploy entire applications. So what should a full-stack framework look like when half its users are machines? In this talk, Sébastien shares how Nuxt is evolving for an agent-first world: conventions that make apps legible to AI, structured logs agents can act on, Markdown pipelines built for LLMs, and tools that let agents inspect a running app. You’ll leave with a clear picture of where Nuxt is heading, and practical patterns to make your own Vue apps easier for both humans and agents to build.

Nitro started as Nuxt’s server engine and became the deploy-anywhere foundation for a whole generation of frameworks. Nitro v3 is a ground-up rethink built on web standards, H3 v2, Rolldown and Vite. Pooya takes you inside the engine: why Request and Response replaced framework-specific abstractions, how one codebase targets Node, edge workers and serverless without changes, and what the new architecture unlocks for Nuxt 5 and beyond. Expect live code, real benchmarks and an honest look at the breaking changes and why they were worth it.

Every AI feature you ship, and every agent that touches your codebase, opens doors attackers are already walking through. Prompt injection hidden in user content. Compromised npm packages slipping into your build. Agents holding tokens with far too much power. Ramona walks through the real threats facing Vue and Nuxt apps in 2027, with live demos of how they’re exploited and concrete defences you can apply the next morning. These cover dependency hygiene, safe rendering of AI output, scoped auth for agents, and security checks that run before code reaches production.

LLMs are non-deterministic, but your UI can’t be. Streaming responses, tool calls, retries, cancellations and partial failures quickly turn an “AI chat feature” into a tangle of boolean flags and race conditions. Maya shows how modelling AI interactions as explicit state machines makes them predictable, testable and easy to reason about. You’ll see how to combine Vue’s reactivity with a state machine to build AI features that handle every edge case gracefully, and why this approach makes your product more reliable than any prompt tweak.

The terminal is having a renaissance. Coding agents, dev tools and AI CLIs all live there, and they need rich, interactive interfaces. What if you could build them with the Vue you already know? Eduardo shows how Vue’s custom renderer API lets you write components, reactivity and composables that render to the terminal instead of the DOM. From layout and keyboard input to live-updating output, you’ll see how to build a polished, agent-style CLI with Vue. Along the way, you’ll learn what renderers reveal about how Vue really works.

Debugging used to mean a developer, a browser and a DevTools panel. Now an AI agent might be the one asking “why did this component re-render?” Guillaume, creator of Vue DevTools, explores how inspection tooling is changing: exposing component trees, state and performance data as structured information both people and agents can query. He looks at what makes a Vue app easy to debug, from naming and component boundaries to observability hooks, and gives a look at where Vue’s developer tooling is heading next.

Writing stories in Storybook has always felt slightly un-Vue: CSF objects, render functions and boilerplate. Not anymore. Julien introduces full Vue SFC support for stories, so you write them the way you write components, with <script setup>, templates and slots, much as Histoire pioneered. He then shows the agentic workflows coming in Storybook 11: AI that understands your component library, generates stories and catches visual and accessibility regressions. You’ll leave with a faster path to a living, tested design system.

Accessibility is no longer a nice-to-have in Europe. It’s a legal requirement for a large share of digital products. This talk translates the European Accessibility Act into concrete Vue practice: accessible component patterns, focus management in SPAs, announcing dynamic content and AI-streamed responses, and catching issues automatically in development with Nuxt’s new accessibility tooling. A practical guide to shipping compliant, genuinely usable apps without slowing your team down.

Vue 3.6’s Vapor mode promises smaller bundles and faster updates by compiling templates straight to DOM operations. But what happens when you adopt it in a real, large codebase? This case study follows one team’s journey: choosing which pages to migrate first, running Vapor and VDOM components side by side, the APIs that didn’t make the jump, and the before-and-after numbers on bundle size, memory and interaction latency. An honest, data-driven guide to deciding whether, and where, Vapor is right for your app.