Beyond Components: Designing Generative UI for MCP Apps
summary
Claim: a model prompted to rewrite a Postman blog produces a search box with a blur animation and accessibility without being asked, so the model now writes better frontend code. The talk maps three concrete approaches: static components (AG UI, Goose), declarative descriptors (JSON/YAML, Vercel JSON Render) and full generative UI where a weather agent returns HTML CSS JS in a single tool call. Trust requires containment, hence MCP apps with the double iframe default as the delivery sandbox.
AI-written summary · about
chapters (13)
- 0:00Introduction: The evolution from 'poor man's bit coding' to high-fidelity UI generation
- 2:56Why are we still stuck with static UI?
- 3:33The new computer: Searching for the interface of the future
- 4:47The role of MCP apps and 'Super Apps'
- 5:52Three levels of UI generation: Static, Declarative, and Generative
- 6:03Understanding Static UI components (e.g., AG UI, Goose)
- 7:42The benefits of Declarative UI (e.g., JSON/YAML renderers)
- 10:06Moving to the next level: Generative UI components
- 11:25The challenge of trust: The need for sandboxing and containment
- 12:22Why MCP apps are the ideal delivery mechanism for Generative UI
- 13:21The 'TV/Radio' analogy: Imagining the future of agent interaction
- 14:46Beyond components: Towards true human-agent collaboration
- 16:16Conclusion: Shaping the future of user interfaces
related talks
21mBeyond Components: The Future of AI UI
29mMCP Apps: Primitives, discovery, and the Future of Software
38m🔥 Generative UI: The Future of Frontend (And Why You Should Care)
12mYour UI Has Left the Building: Interfaces That Write Themselves
33mFrontend after AI: The new UX
51mBuilding Better Components — A Journey to a More Resilient Frontend
38mEvolving AI chat with MCP Apps
05mGenerative UIs