From idea to deployed app, without leaving the page
Describe what you want, watch it build in a live preview, and ship it — all in one tab. Here's how a single sentence becomes a running app.
Blog
Notes from the workshop.
How we think about building software with AI — the model choices, the guardrails, the infrastructure, and the small decisions that add up to a tool we’d actually want to use.
Built-in web search across Claude, GLM, GPT, and Gemini
Your coding agent can look things up while it works — current docs, API shapes, package versions — using each provider's own server-side search.
A private machine for every project
Every project runs in its own isolated virtual machine — not a shared container, not a browser tab. Here's why that isolation is the foundation.
Plan before you build: how plan mode works
Before touching a single file, the agent can lay out exactly what it intends to do — so you approve the approach, not just react to the result.
Why we let you pick the AI for each step
Claude, GLM, GPT, or Gemini — and an Auto mode that routes for you. No single model is best at everything, so we don't pretend otherwise.