Reconstruct

npx add-mcp 'https://reconstruct.dev/mcp'

Reconstruct AI skills

Reconstruct captures web interfaces — screenshots, element trees, styles, assets, routes and recorded flows — and serves them to AI coding agents through a hosted MCP server. This repository packages that server together with nine agent skills that guide an agent from inspecting a reference interface to building, verifying and launching a product.

It works as a Claude Code plugin, a Cursor plugin, a Gemini CLI extension and a plain Agent Skills collection.

Install

Claude Code

/plugin marketplace add cryptify-corporation/reconstruct-ai-skills
/plugin install reconstruct@reconstruct

Run /mcp, select reconstruct and sign in with your Reconstruct account in the browser.

Any agent with Agent Skills support

Install the skills with the skills CLI for Claude Code, Codex, Cursor and other agents:

npx skills add cryptify-corporation/reconstruct-ai-skills

Add -a claude-code, -a codex or another agent name to choose the target, and --skill '*' to select all skills. The same catalog is published at https://reconstruct.dev, so npx skills add https://reconstruct.dev installs identical files. The skills need the Reconstruct MCP server; connect it as described in Connect AI agents.

Cursor

Install Reconstruct from the Cursor Marketplace. To try it before then, clone this repository into ~/.cursor/plugins/local/reconstruct and run Developer: Reload Window.

Gemini CLI

gemini extensions install https://github.com/cryptify-corporation/reconstruct-ai-skills

MCP server only

The server is available at https://reconstruct.dev/mcp over Streamable HTTP with OAuth sign-in. In Claude Code:

claude mcp add --transport http reconstruct https://reconstruct.dev/mcp

Guides for Claude Desktop, VS Code, Codex, Cursor, Gemini CLI and other clients, including personal access tokens, are in the documentation. The server is also listed in the MCP Registry as dev.reconstruct/reconstruct.

Skills

Skill	Purpose
reconstruct-inspect	Inspect a captured interface and plan a bounded reconstruction task
reconstruct-design	Extract measured design tokens and component specifications
reconstruct-build	Implement a component, page or recorded flow in your existing stack
reconstruct-verify	Compare the result with pinned capture evidence: functional, visual and accessibility checks
reconstruct-backend	Design or implement API contracts, persistence, authorization and integrations behind the UI
reconstruct-entrepreneur	Research demand and differentiation from real user reviews and competitor complaints
reconstruct-brand	Create naming, visual identity and voice; replace reference branding
reconstruct-launch	Prepare positioning, landing pages, pricing proposals and store listings
reconstruct-deploy	Prepare and verify a release: hosting, DNS, migrations and production configuration

Each skill works on its own. Supporting templates live inside the skill's directory.

Example

Use Reconstruct to inspect the pricing page of <product>, then build it as a React component in this project and verify it against the capture.

More prompts: prompt examples.

Access and data

Reading captures requires a Reconstruct account; the plugin never asks for your password and signs in through your browser. Captured websites are treated as untrusted data. The skills submit new capture requests only when you ask for one. See the privacy policy and terms.

Support

reconstruct.dev/support or open an issue in this repository.

License

Reconstruct AI Skills are part of Reconstruct, a product owned by Cryptify. Copyright (c) 2026 Cryptify. All rights reserved.

These skills are provided under a proprietary license for personal or internal business use with Reconstruct. Redistribution, resale, and publication of modified versions require prior written permission from Cryptify. See LICENSE. Third-party attribution and retained upstream permission notices are in NOTICE.