Loading patterns…
Generative UI (Agent-Rendered Interfaces)(GenUI)
The agent generates, selects, and controls the interface itself at runtime, returning rich interactive components (forms, charts, dashboards, maps, multi-step widgets) instead of plain text, and deciding HOW to present a result from the content type and the user intent. A model tool call emits UI that the host renders in a sandboxed frame, for example MCP Apps serving bundled HTML through `ui://` resources, or portable specs like Google A2UI and Open-JSON-UI that the client paints natively. This crossed from experimental to production in 2026 as MCP Apps shipped across Claude, ChatGPT, VS Code, and Goose, alongside A2UI and CopilotKit/assistant-ui generative UI. Distinct from `ag-ui-protocol`, which streams agent EVENTS to a prebuilt frontend, and `adaptive-interface-patterns`, which personalizes an EXISTING interface: here the agent materializes the interface itself, choosing and constructing the components at response time.
In 30 seconds
- What
- Agent generates and returns interactive UI components (forms, charts, tables, maps) as tool outputs, which the host renders in sandboxed frames.
- When to use
- Results benefit from visual presentation, user needs to interact with data, or the best format depends on intent and content type.
- Watch out
- Rendering failures, network latency, and browser compatibility issues compound agent errors; test UI generation paths thoroughly before production.
Ask the AI expert about this pattern
Opens the assistant with your question prefilled. You review it before sending.
Generative UI (Agent-Rendered Interfaces): Overview
The agent generates, selects, and controls the interface itself at runtime, returning rich interactive components (forms, charts, dashboards, maps, multi-step widgets) instead of plain text, and deciding HOW to present a result from the content type and the user intent. A model tool call emits UI that the host renders in a sandboxed frame, for example MCP Apps serving bundled HTML through `ui://` resources, or portable specs like Google A2UI and Open-JSON-UI that the client paints natively. This crossed from experimental to production in 2026 as MCP Apps shipped across Claude, ChatGPT, VS Code, and Goose, alongside A2UI and CopilotKit/assistant-ui generative UI. Distinct from `ag-ui-protocol`, which streams agent EVENTS to a prebuilt frontend, and `adaptive-interface-patterns`, which personalizes an EXISTING interface: here the agent materializes the interface itself, choosing and constructing the components at response time.
- Agent picks the presentation: a chart for a trend, a form for an action, a table for a comparison
- Tool calls return UI components, not just text, which the host renders inline
- Sandboxed rendering: MCP Apps serve bundled HTML via `ui://` resources in an isolated iframe
- Portable UI specs (Google A2UI, Open-JSON-UI) painted natively by the client, host-agnostic
- Bidirectional: rendered widgets post user actions back into the agent loop as tool results
- Content-and-intent routing decides layout, density, and modality per response
Get the Agent UX field guide
All 21 agent-UX patterns condensed into one guide: what each is, when to reach for it, and what bites in production. The link arrives with your confirmation, alongside the weekly Agent Architect.
Weekly email, one-click unsubscribe. We only use your address to send the briefing.
References
The papers, specifications, and repositories this pattern is based on.
From the engineer behind this catalog
Get your agent architecture reviewed
This page documents one pattern. Your system runs dozens, and most failures live in how they fit together. Have the whole design reviewed against the 288 patterns in this catalog: architecture, reliability, evaluation and cost, every finding mapped to the pattern that fixes it.
€750 instead of €1,500, one week, written report and walkthrough call, until 30 September