HeyBeauty MCP Server

Overview

HeyBeauty MCP Server is a TypeScript‑based Model Context Protocol implementation that bridges AI assistants with the HeyBeauty virtual try‑on API. It solves a common pain point for developers building fashion or e‑commerce experiences: how to let an AI assistant seamlessly fetch garment data, initiate a try‑on computation, and retrieve the resulting visual output without writing custom integration code. By exposing garments as lightweight resources and providing tools to submit and query try‑on tasks, the server abstracts away authentication, request orchestration, and state management.

The core value lies in its ability to turn a simple image URL and garment description into an enriched prompt that an LLM can use to generate realistic try‑on images. Developers can reference garments through URIs, automatically pulling in metadata such as name, description, and image URLs. The tool stores the task in server state and triggers the external API, while lets the assistant poll for completion. The dedicated prompt aggregates all required inputs into a structured format, ensuring consistency across different LLMs and use cases.

Key capabilities include:

• Resource abstraction – garments are represented as first‑class resources with unique URIs, making them discoverable and reusable across multiple assistants.
• Tool orchestration – two robust tools handle task submission and status retrieval, allowing the assistant to manage long‑running operations transparently.
• Prompt templating – the prompt provides a ready‑to‑use format that includes user image, garment image, and descriptive metadata.
• Stateful server design – task data persists in the MCP server’s internal state, enabling retries and history tracking without external databases.

Typical use cases span virtual fitting rooms for online retail, personalized styling recommendations in fashion apps, and creative AI demos that showcase garment transformations. In a workflow, a user uploads a selfie; the assistant references a resource for a dress, submits a try‑on task via the tool, and then streams back the rendered image once completed. Because all interactions are defined by MCP, any client that supports the protocol—Claude Desktop, web agents, or custom front‑ends—can integrate this functionality with minimal effort.

What sets HeyBeauty MCP Server apart is its tight coupling to the HeyBeauty API while remaining agnostic to the LLM backend. This means developers can swap out Claude for another model without changing server logic, and they gain a reusable virtual try‑on pipeline that can be extended with additional resources or tools as the product evolves.