Gradio

Gradio is a Python library for putting a web interface in front of a function, most often a machine learning model, without writing frontend code. Hugging Face acquired it in 2021, and Hugging Face Spaces is the hosting its docs point to first.

How it works: you hand Gradio a Python function plus input and output components, and it builds the page. gr.Interface does that for a single function, gr.ChatInterface builds a chatbot UI around a chat function, and gr.Blocks gives you custom layouts, multiple data flows and event listeners when one function isn't enough. demo.launch() starts a local server.

What you get: every app also exposes an API, with a "Use via API" page and official Python and JavaScript clients; mcp_server=True turns those endpoints into MCP tools for LLMs. A built-in queue streams progress over Server-Sent Events. share=True creates a temporary public link that expires after a week, and you can deploy to Hugging Face Spaces, embed an app with a web component or iframe, or mount it inside a FastAPI app with gr.mount_gradio_app. Password auth and Hugging Face OAuth are built in.

The limits: styling goes through themes, css= and js=, but Gradio warns that query selectors in custom CSS and JS aren't guaranteed to keep working across versions because the DOM can change. The queue runs one worker per event by default, so concurrency for heavy functions is something you tune. And despite "free" in Gradio's own docs, Hugging Face's Spaces docs now say Gradio Spaces need a paid plan, apart from up to two ZeroGPU Spaces on a free personal account.