Poe, Quora’s platform that integrates multiple AI models, has introduced a new feature that enables users to build visual interfaces, essentially apps, on top of any combination of models.

This feature, called Poe Apps, allows users to describe the app they want to create using the App Creator tool. The descriptions can include specific models, such as OpenAI’s o3-mini or Google’s video-generating Veo 2, or more general specifications.

The App Creator tool, powered by Anthropic’s Claude 3.7 Sonnet, translates the description into code for the app interface and custom logic expressed in JavaScript. Poe Apps can run alongside the Poe chatbot window or be entirely visual, and the underlying code is exposed for manual adjustment and fine-tuning.

Quora has created example apps, including one that transforms photos into 3D anime-style art using OpenAI’s GPT-4o and Black Forest Labs‘ Flux-Pro-1.1. Another example app removes unwanted objects from images, utilizing Bria’s Bria-Eraser model.

Poe Apps can be shared with other Poe users, currently only on the web, with iOS and Android support forthcoming, according to Quora. Each time an app uses an AI model, it draws from the user’s point balance with Poe. Free users receive a daily point allotment, while premium tier subscribers, at $9.99 per month, get flexible daily or monthly point packages.

In a blog post, Quora, noting that App Creator is available at a reduced early-access price for a limited time, hinted at potential app monetization options.

“This is an early launch, and we have a long roadmap ahead to give creators more power, including the ability to earn money directly from their apps,” the company wrote. “We are excited to see what you all create, and we are excited to see how much better we can make Poe Apps as the models continue to improve at writing code this year.”

Poe Apps, building on the web apps feature Poe launched last July, are similar to Anthropic’s Artifacts and OpenAI’s ChatGPT Canvas tools, which provide dedicated workspaces for editing and adding to AI-generated content like code and documents. Although the apps produced by these tools have limitations, they demonstrate the significant progress made in models’ programming capabilities.