Generative UI
Solves: trusting controls an agent built on the spot inside your product.
Emerging — seen across multiple products; the shape is still settling.
Generative UI is interface an agent composes on the spot. The person can tell agent-built controls from the product's own. The controls reach only what they were allowed to, and their origin is marked.
When to Use It
- Agents compose UI at runtime inside your product
- Third-party surfaces need sandboxing and a declared contract
- Generated interface must be distinguishable from host interface
When Not To
- A fixed template would do — generation adds risk without value
- You can't sandbox rendering — don't ship what you can't contain
- Generated UI would collect input the host never sees
Three views
The exchange between the human, the agent and the system the agent acts on; who holds each part of it; and the component that implements it.
- 01 Slot The place on screen where AI-built controls appear
- 02 Sandbox The fence that keeps AI-built controls from reaching anything else
- 03 Allowlist The parts the AI is allowed to build with
- 04 Provenance A mark that says the AI made this control
generative ui · primitives
primitive wireframe, generated from the anatomy — the installable component ships when this pattern's anatomy stabilizes
Field Notes
Documented sightings (3) — anatomy-mapped observations from vendor materials
The revised agent dynamically shifts its reply style — greeting and affect-reception vs. direct execution — based on inferred user stance, constituting an adaptive response surface.
Contract-first generated UI: servers declare surfaces, host sandboxes rendering, provenance built into the protocol.
Inline apps: third-party UI renders in conversation; host controls invocation and rendering scope.
Side by Side
| Anatomy part | Anthropic | OpenAI |
|---|---|---|
| Slot | declared UI surface in host | inline app surface in chat |
| Sandbox | host-enforced isolation | host-controlled rendering |
| Allowlist | declared surface contract | app review process |
| Provenance | server identity in protocol | — |
What Vendors Call It
Chart unlocks as the corpus grows — n = 0 mentions across 0 quarters.
- Anthropic MCP Apps / artifacts
- OpenAI Apps SDK / canvas
Tensions & Failure Modes
- Expressiveness vs. safety: the more the agent can render, the more it can misrender
- Host coherence: generated UI that ignores host conventions erodes trust
What we are learning from the community
Readers who build with Generative UI add what they found: questions, use cases, critiques, other ways to solve it. An editor reads each one before it shows here. Nothing here changes the pattern's status on its own.
This pattern has no contributions yet. The form below sends the first one.
The Story So Far
- Aug 30, 2026 Seeded from SPEC v1.5 Phase 0: Anthropic-first corpus with OpenAI counterparts
Related Patterns
The Agents & Humans Briefing
Agentic experience design, coding agents, MCP, and the signals that matter — weekly, free, in about five minutes.
Free. No spam. Unsubscribe anytime.