Bubble merges AI prompting with drag-and-drop building

Coding with AI via plain English feels magical right up until you try to build something meant for production. One wrong prompt and the entire codebase breaks, leaving non-technical builders stuck staring at an inscrutable wall of code.

No-code giant Bubble is stepping in with a hybrid approach. The company just rolled out Bubble AI Agent, a new tool designed to blend natural language coding with drag-and-drop visual development for production-grade web apps.

Chat to build, drag to refine

A recent survey of Bubble users revealed a sharp divide: while AI excitement is everywhere, only 9% feel comfortable using AI coding tools for business-critical applications, and 72% actively worry about security vulnerabilities in AI-generated code.

Bubble's answer is not to replace the canvas, but to anchor the AI to it. Users can describe what they want in natural language to generate functional modules, and then immediately switch over to Bubble’s visual editor to manually tweak layouts, database structures, or backend workflows.

Bubble co-CEO Emmanuel Straschnov noted that builders want AI speed without compromising production stability. The hybrid model keeps the code editable and fully transparent.

To drum up interest, Bubble teamed up with Anthropic to launch a $25,000 hackathon challenging developers to test the agent. The company also teased several upcoming features for later this year, including an AI-enhanced Property Editor, native in-app purchases for mobile apps, and the full integration of its security dashboard (formerly Flusk).

Bubble has raised $106 million to date from backers like Insight Partners and SignalFire.

Why it matters

The novelty of one-shot AI app builders has largely worn off. For solo founders and early-stage teams, the real headache begins at day two, when an app needs maintenance, custom business logic, and security patches. By tethering an AI agent to a deterministic visual framework, Bubble gives users a visual escape hatch when the AI hallucinates, making production-grade deployment much less risky.

We think this is the right trajectory for AI-assisted development, but the real hurdle will be execution. Visual platforms already carry their own architectural complexity; adding automated prompting into the mix could easily create hard-to-debug logic errors inside visual workflows. Whether its integrated security tooling can truly alleviate user anxiety will decide if businesses actually trust it with critical apps.