Now
Overview
MDS is a language. People learn its vocabulary, grammar, patterns and judgement over time, and AI tools are now learning it too. We’ve already started with experiments: teaching it through skills and knowledge, and surfacing that learning where people work via the MCP server, the MDS AI Agent and Figma Make Kit in MDS AI. Q4 and into the new year is about taking both the teaching and the channels from experiment to production, so AI can work with MDS well enough to produce code and designs that are consistent, accurate and trustworthy.
AI support is the priority
Four pieces of work make up our MDS AI support focus: three improve what AI learns about MDS, and one improves how that learning reaches the places people work.
- Component documentation quality: bring every component page to a High quality rating, so the vocabulary and grammar are clear and consistent for AI and people alike. See Content Quality Dashboard
- API descriptions quality: fix recurring problems in property, event and slot descriptions across all packages, so every term means one thing and agrees with the component behind it. See API Descriptions Quality Dashboard
- How our knowledge is organised: understand how people use MDS with AI, then organise our documentation and guidance so the patterns and structure match what they need to produce
- How our knowledge reaches AI tools: check that the foundations behind our MCP server, MDS AI Agent and other AI endpoints are fit for purpose, and optimise how MDS works in Claude Design, Claude Code and GitHub Copilot, cross-checking with the Figma Make Kit, so MDS can be “spoken” wherever people already work
Product features
Alongside the AI work, we are also progressing a number of product features across MDS:
- Chat component: our component for building AI conversation experiences, moving from experimental to supported production
- AI colour role: experimental design tokens and research into how AI-generated content is labelled across the business
- Component focus state: a new focus state that works in any layout and fits the refreshed brand
- Rem support: the experimental feature is built, so we are now bringing it to users. It lets products built with MDS scale with each end user’s browser font size
A long-requested feature
Data visualisation gets official, non-experimental support, something our users have asked for over many years. It brings design tokens, Figma components, charting library configuration and rewritten guidance, replacing guidelines that have drifted from our visual language.