GitHub open-sources Replica: AI prototyping with your own design system

What it's about

On October 28, 2026, GitHub released Replica as an open-source project. The tool comes from GitHub Design, the company's design team. It's meant to help turn ideas into interactive prototypes quickly, meaning clickable mockups of a software interface. The code is under the MIT license, which allows free use and modification. Replica is not an AI model with downloadable weights, but a tool that uses AI through the GitHub Copilot app.

What Replica does differently

GitHub sets Replica apart from general AI prototyping tools, which often produce generic interfaces. Replica instead builds on what the team already has: existing components, patterns and documentation, in other words the design system. A design system is the shared collection of building blocks such as buttons, forms and rules that keeps a product looking consistent.

Roles are split: designers set the guidelines. Everyone else on the team, such as product managers or developers, can then describe in natural language what they want to try. They see, test and refine the result alongside their conversation with Copilot.

What you can do with it

GitHub lists four use cases:

• Prototype with your design system: Existing components and documentation steer the AI-generated drafts.
• Make ideas interactive: You describe what you want to explore and refine the result in conversation.
• Collaborate and gather feedback: Prototypes can be published to an optional, self-hosted Replica Hub, where colleagues can view and comment on them.
• Adapt it to your team: The tooling and guidelines are open and can be tailored to your product and way of working.

According to GitHub, teams should get from conversations to tangible experiences faster, test assumptions and make better-informed product decisions.

Availability and getting started

Replica is available now as an open-source project. You can start locally or set up the Replica Hub to share prototypes within your team. Setup instructions, documentation and examples are in the GitHub repository. GitHub invites the community to share feedback, ideas and contributions. The source mentions no pricing, no exact system requirements and nothing about which Copilot plans are required. Those points remain open.

Takeaway

The approach is interesting because AI prototyping often falls short by having no connection to the real product. If the AI works with your own components, prototypes can come closer to reality. How well that works in practice is unclear, since the source is a vendor announcement with no tests or benchmark figures. It's also unclear how much upfront work defining the design guidelines takes, and whether Copilot is strictly required. For product teams with an established design system, it's worth a look. For individuals without a design system, the benefit is less clear.