What do I need to start?

A sentence is enough. You can also bring references and screenshots, connect a GitHub repository, or capture a live web app you own or have permission to analyse. The Explore plan is free and needs no card.

How does the design agent charge?

In usage units from your plan's monthly allowance. Requests up to your workspace's auto-approve limit run straight away; the limit is 5 units unless someone who manages billing changes it, and 0 means always ask. Anything bigger shows a quote first, and you're charged only what the work used. Generating images and video spends credits instead of units.

Will M1rror change my code?

Not on its own. Scans and pull request checks read your repository without running it. Code changes reach your repository only as a draft pull request that your team reviews and merges. M1rror never pushes to your default branch, and never approves or merges.

What does it work with today?

Code features support React and TypeScript web apps. Screens are designed at phone (390 × 844), tablet (768 × 1024) and desktop (1280 × 800) sizes, and design systems export as DTCG tokens, CSS variables, TypeScript, JSON, SwiftUI, Android resources and Jetpack Compose.

Do reviewers need a paid seat?

No. Reviewers can comment and approve on every plan without a paid seat, and guests can review through a private link without an account. Only editors, who change systems and screens, use seats.

Is my work used to train AI models?

No. Your repositories, references and captures are private to your workspace by default and are not used to train AI models.

Which coding agents work with M1rror?

Any agent that supports remote MCP servers. Settings › Connect your agent has ready-made setup for Claude Code, Claude Desktop, Cursor and Codex. Agent connections are included on the Studio and Team plans.

Frequently asked questions

Answers about getting started, how the design agent charges, whether M1rror changes your code, what it supports, reviewer seats, AI training and what isn't available yet.

These are the questions people ask most. If yours isn't here, use Send feedback in the app's Help menu.

Questions

What do I need to start?: A sentence is enough. You can also bring references and screenshots, connect a GitHub repository, or capture a live web app you own or have permission to analyse. The Explore plan is free and needs no card.
How does the design agent charge?: In usage units from your plan's monthly allowance. Requests up to your workspace's auto-approve limit run straight away; the limit is 5 units unless someone who manages billing changes it, and 0 means always ask. Anything bigger shows a quote first, and you're charged only what the work used. Generating images and video spends credits instead of units.
Will M1rror change my code?: Not on its own. Scans and pull request checks read your repository without running it. Code changes reach your repository only as a draft pull request that your team reviews and merges. M1rror never pushes to your default branch, and never approves or merges.
What does it work with today?: Code features support React and TypeScript web apps. Screens are designed at phone (390 × 844), tablet (768 × 1024) and desktop (1280 × 800) sizes, and design systems export as DTCG tokens, CSS variables, TypeScript, JSON, SwiftUI, Android resources and Jetpack Compose.
Do reviewers need a paid seat?: No. Reviewers can comment and approve on every plan without a paid seat, and guests can review through a private link without an account. Only editors, who change systems and screens, use seats.
Is my work used to train AI models?: No. Your repositories, references and captures are private to your workspace by default and are not used to train AI models.
Which coding agents work with M1rror?: Any agent that supports remote MCP servers. Settings › Connect your agent has ready-made setup for Claude Code, Claude Desktop, Cursor and Codex. Agent connections are included on the Studio and Team plans.
What isn't available yet?: Live previews and a built-in Mobbin catalog are switched off for now, and Figma sync depends on how a deployment is set up. During the pilot, M1rror runs builds itself only on the sample project; for your own repository, your coding agent builds from M1rror's approved plan over MCP. Where something isn't available, the product says so.

Last updated 5 November 2026

Edit this page on GitHub