Design System MCP: Delivering context where the work happens

Aparat, a video-sharing service, built a server that AI coding agents query for its design-system rules, so developers stopped copying outdated files into their agents.

Ehsān Vaeghi works on the design systems at Aparat, a video-sharing service with more than 60 million monthly active users. Aparat keeps three design systems, for the web, its phone apps and smart TVs, and writes each component's behaviour into structured files for AI agents. Those files reached seven development teams by hand, through chat messages and local copies that went out of date.

• The server uses MCP, the Model Context Protocol that AI coding tools use to call outside services, and updates within seconds of each change to the design systems.
• Each developer or team gets a key scoped to one or more design systems, so a TV developer's agent is refused the app and web systems.
• Every request records the version of the rules that answered it, so the design-system team can warn teams that last read an older version.
• Every component or screen an agent looks for and cannot find is logged, and the team treats that list as its backlog.

Vaeghi argues that a design-system team should serve its rules to AI agents rather than hand over files, so the rules stay current and their use is visible. The piece reports usage tracking but no measured effect on the code teams ship.

Derived by Working Surface from the article; more in the Playbook. Source line: If context is what makes the difference, and the design system team is its owner, then we shouldn’t hand context over (and lose control) at all.

14 October 2026: as AI agents raise the volume of code changes, teams are moving routine checks into automated tests and installable rules, and keeping people for intent and contested decisions.