Montage (M1)

Montage is a runtime framework for agentic user interfaces. Instead of having an LLM regenerate every chart, table, kanban, or form token-by-token each turn, agents emit a small intent schema and Montage compiles it server-side into production components. The team reports 10x faster hydration and 50–100x fewer output tokens compared to token-streamed UI, with model-agnostic and framework-agnostic output themed to the host brand. Aimed at companies shipping customer-facing AI agents that want consistent, branded interfaces without the per-turn inference cost of full UI regeneration.