Chat component and intent-driven documentation
Continuing the AI focus and pace we have built in Q2, and following the launch of MDS AI plus our Agent and Figma Make support, we are introducing two important MDS updates.
Chat component
We have shipped a new experimental Chat component.
The goal is simple: help teams add chat experiences without building everything from scratch.
What it provides
- A simple framework for a complex interaction pattern
- LLM- and AI-platform-agnostic setup
- Core chat experiences out of the box
- Flexible hooks for custom behaviour and integrations
This follows the same approach as our Table component. We provide a solid foundation, and teams extend it to match their needs.
Current status
- Available in code only
- Already in use on the MDS website, powering MDS Chat below
What’s coming next
We are actively developing this component, and new features are coming very soon, including:
- Built-in and customisable response actions, for example like and dislike
- More layout options, inline page and drawer variants, alongside the current popup
- Reasoning output
- Chat component guidance and a revisit of our Conversational AI guidelines
- Designer support in Figma
Strengthening the semantic backbone of MDS
A key part of our updated vision is making MDS guidance clearer, more consistent, and usable for both humans and AI agents.
We have started this work by updating the Button guidelines.
What’s changed
- A stronger focus on intent and usage
- Clearer rules to reduce ambiguity
- More structured guidance that can be interpreted programmatically
What’s next
- Rolling this structure into Agent instructions and skills
- Standardising the approach across all component documentation
- Tightening broader UX guidance to align with this direction
As always, we’d love your feedback as we continue to shape this work.