CopilotKit Releases OpenIntelligentUI, an Open-Source Alternative to ChatGPT Intelligent UI
·2026.10.21 09:30
Key point
The MIT-licensed project, which reached 3.6k GitHub stars and Trending #5, allows agents to dynamically switch between text, native components, and custom interactive UIs.
Details
- CopilotKit released OpenIntelligentUI, an open-source (MIT) implementation of ChatGPT's Intelligent UI that converts user queries into explanations, comparisons, or working tools.
- The project achieved 3.6k GitHub stars, 397 forks, and ranked #5 on GitHub Trending and #3 on Trendshift for TypeScript repositories.
- An agent named Jev selects the response format for each turn, routing basic tables to A2UI and complex visualizations like charts, maps, and calculators to Open Generative UI.
- The system uses a Python Deep Agent backend and a Next.js frontend, streaming custom HTML/CSS/JavaScript into isolated iframes via the generateSandboxedUi tool.
Architecture and Functionality
- Three primary modes: Understand (step-by-step mechanisms), Compare (side-by-side trade-offs), and Make a tool (interactive controls like bill splitters or 3D models).
- Visualization routing is deterministic; if a provider fails, the system surfaces the error rather than silently switching to a different model or router.
- Custom UI parameters follow a specific order: initialHeight, placeholderMessages, css, html, jsFunctions, and jsExpressions.
Setup and Requirements
- Requires Node.js 22+, pnpm 9+, Python 3.12+, and uv for local execution.
- Users must provide their own OPENAI_API_KEY for the chat-latest model and TYPESAFE_API_KEY for the Jev agent.
- API keys are stored only in browser memory and are not persisted to storage or included in chat state checkpoints.
This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.
Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.