Agent Builder by Thesys - Build AI agents that respond with UI instead of text
by•
Build AI agents that reason dynamically and respond with charts, cards, forms, slides and reports. No workflows. No code. Just connect your data, add instructions, customize style, and publish and share with anyone or embed on your site.
Replies
You’re growing fast, congratulations on the new launch!
Thesys
@mykyta_semenov_ thanks for the support
You mention C1 handles UI generation and runtime under the hood that’s super interesting.
From a builder perspective, I’m curious:
is the rendering primarily client-side, server-driven, or hybrid?
And when generating something like a multi-series chart during a live conversation, how do you keep the experience responsive?
Really excited about the idea of LLMs returning structured, interactive UI instead of just text.
Thesys
@rachid_jeffali
There are two parts to this - UI generation and UI rendering.
UI generation is done by the AI Agent.
In the current implementation, the actual rendering is performed on the client side. Our architecture decouples UI generation from rendering, allowing us to support multiple platforms.
The maker mentioned moving from regenerating charts per turn to "live charts where data can be refreshed." thats a fundamentally different architecture problem, stateless generation vs stateful component lifecycle. How is C1 handling component identity and diffing between turns today?
This is interesting because it shifts the problem from “better prompts” to better outputs. Generative UI feels like the missing layer between LLMs and real products — text is rarely the final interface users want. Curious how you’re handling state + re-renders when responses evolve over time.
How does the No-code builder handle complex logic or multi-step workflows compared to the C1 API?"
How do you see Generative UI evolving compared to traditional static dashboards in the next year?
Thesys
@quang_tri_u The thing about static dashboards, everyone person in the team wants different static dashboards for the same data.
Teams can ship faster with GenUI rather than building dashboard for each usecase.
The pivot from developer API to no-code agent builder is a smart distribution move, but I'm wondering how you handle the tension between the no-code audience and the c1 api developer audience are these the same buyers or are you splitting focus?
Thesys
@leadsgenerationbooster We started as a developer-first API platform, and that remains core to our strategy. Agent Builder came from demand by non-technical founders and product managers who wanted a faster way to experiment with GenUI agents.
We don’t see this as splitting focus. It’s the same platform with two entry points: the API for developers who want full control, and Agent Builder for teams who want speed and accessibility.
A01
congrats on the launch!
This is a big shift — agents that return usable UI instead of paragraphs feel way more practical. Curious how this changes how we design AI apps going forward 👀