Sketch launches Plugin API v4 with real-time model inference support

Design · 4 min read

Plugin API v4 introduces a first-class inference endpoint inside Sketch that lets plugin authors connect to models running locally or in the cloud, with lifecycle and permission controls managed by Sketch. This removes previous workarounds where plugins had to implement separate microservices or rely on external tooling.

Plugin authors can now create features like live copy suggestions, instant component refactoring, or accessibility scanning with near-zero perceived latency. Sketch has also published a set of sample plugins demonstrating safe model usage patterns and UI affordances for displaying model uncertainty.

To address privacy and security, v4 includes a permission model where users explicitly allow data sharing per plugin with scoped session tokens. Sketch says the update will accelerate a new generation of intelligent plugins while keeping user trust central to the platform.