Figma updates Plugin API with live model pipelines and token quotas

Tech · 4 min read

Figma announced an overhaul to its Plugin API enabling persistent model sessions, streaming outputs, and centralized usage quotas tied to organization billing. Plugins can now maintain authenticated WebRTC connections to inference endpoints, minimizing latency for iterative prompts and live previews.

The update also introduces token quotas and deterministic rate limits so admins can control generative spend at the team level. Figma provides new UX hooks for plugins to show provenance badges, confidence scores, and to request explicit designer approval before applying bulk changes.

Plugin developers welcomed the clearer cost model and streaming support, which make it easier to implement features like collaborative co-editing with an assistant model. Designers expect fewer surprises in their bills and better visibility into when and how AI changed their files.