Thesys OpenUI is an MIT, self-hostable generative-UI runtime: an agent emits a typed spec and OpenUI renders real interactive components instead of plain text.
Yes, Thesys OpenUI (Open Standard for Generative UI) is open source under the MIT license.
Free and open source (MIT), runs locally; no paid tier or pricing wall is stated in the repo. (free).
Yes, Thesys OpenUI (Open Standard for Generative UI) can be self-hosted (the source is available under the MIT license).
An MIT-licensed, self-hostable runtime for **generative UI** from Thesys Inc.: rather than hand-coding every screen an assistant might need, your agent emits a typed UI specification and OpenUI renders it into real, interactive components — cards, sliders, buttons, forms, layouts — on the fly. The pitch, demoed publicly as generative hotel cards, a budget slider, and live controls, is that a model can return a usable interface instead of a wall of text, and the interface is produced from a declared spec rather than free-form markup. It is published under MIT and runs locally.
As lean teams wire agents into real GTM work — qualifying a lead, configuring a quote, booking a slot, triaging a reply — the bottleneck stops being the model's answer and becomes the interface: a text-only agent forces the user back into copy-paste, and hand-building a bespoke screen for every agent action does not scale. Thesys OpenUI is one answer: the agent declares a typed UI spec and the runtime renders interactive components, so an assistant can present a form or a set of controls as its response. The honest read: the headline "6M downloads" is an unverifiable npm claim, the name is badly squatted so you must install from the canonical thesysdev/openui, and this is a rendering runtime rather than a full agent framework — you still own the model and the logic. What is verified is the part that governs adoption: MIT read from the LICENSE file, self-hostable and free, and a repo that was pushed the same day it was checked.