MCP 服务器

io.github.wubin1836/ai-hive-aigc

AI Hive AIGC MCP server enables commercial image and video generation for e-commerce, marketing, and advertising, with model discovery, media upload, generation tasks, and polling.

README

AI Hive AIGC MCP Server

AI Hive AIGC is a Streamable HTTP MCP server for commercial image, video, e-commerce, marketing and advertising generation. One server exposes model discovery, media upload, image generation, video generation, e-commerce workflows, advertising/TVC workflows and task polling.

AI Hive AIGC MCP 为中文用户和全球开发者提供统一的图片与视频生成入口，覆盖 Nano Banana Pro、GPT Image 2、Seedream、Seedance、MiniMax H3、Wan、HappyHorse，以及电商主图、商品详情页、Listing、PDP、广告、TVC、带货、种草、短剧和漫剧场景。

MCP endpoint

https://ai-hive.iclip.cn/api/mcp

Transport: Streamable HTTP.

Authentication

The public endpoint advertises OAuth 2.1 protected-resource metadata and supports PKCE. The local reference server also supports an AI-HIVE API key in the request header:

x-ai-hive-api-key: sk-api-...

For local development, set AI_HIVE_API_KEY. The local server intentionally allows unauthenticated MCP discovery so registries can index the tool catalog; generation calls fail safely until a user provides a key.

Tools

ai_hive_list_models
ai_hive_upload_media
ai_hive_generate_image
ai_hive_generate_ecommerce_image
ai_hive_generate_video
ai_hive_generate_ecommerce_video
ai_hive_generate_advertising_video
ai_hive_get_task

Generation tools are non-idempotent and can incur charges. After receiving a task_id, query that task instead of resubmitting the same request.

Run locally

npm install
npm test
npm start

Health check: GET /health

Static registry card: GET /.well-known/mcp/server-card.json

The production MCP is hosted by AI-HIVE and is not deployed from this repository. This repository provides the public protocol reference, tool implementation, tests, Registry metadata and policy documents.

Docker

docker build -t ai-hive-mcp .
docker run --rm -p 3000:3000 -e AI_HIVE_API_KEY=sk-api-example ai-hive-mcp

Smithery configuration schema

When publishing, use a session configuration field injected as an HTTP header:

{
  "type": "object",
  "properties": {
    "apiKey": {
      "type": "string",
      "title": "AI Hive API Key",
      "x-from": { "header": "x-ai-hive-api-key" }
    }
  },
  "required": ["apiKey"]
}