Micu Image MCP

Micu Image MCP

Wraps the Micu image API as an MCP server for generating, editing, batch processing, and multi-reference image fusion, supporting GPT-image-2 and Grok models.

Category
Visit Server

README

米醋画图 MCP

米醋 的图像接口包装成 MCP server,让 Claude Code / Codex / Cursor 等 MCP 客户端直接生图、改图、批处理、多图参考。

默认使用 gpt-image-2 / gpt-image-2-pro。可选配置 MICU_GROK_API_KEY 后,也能走米醋 Grok 图像通道,当前实测模型包括:

  • grok-imagine-image-lite
  • grok-imagine-image
  • grok-imagine-image-pro
  • grok-imagine-image-edit

功能

Tool 说明
image_generate 文生图。米醋 image2 支持 1K / 2K / 4K;Grok 支持 1K / 2K 路由
image_edit 单图参考/编辑。image2 走 edits 或 reference_image;Grok 走 reference_image
image_batch_edit 多张图逐张同指令处理
image_multi_reference 2-10 张参考图融合成 1 张新图;Grok 走 image_urls
server_info 查看 base URL、模型、size 规则、重试策略、安全约束

第一次使用前,让 LLM 调一次 server_info,可以看到当前运行时配置和可用能力。


Grok 与 GPT Image2 功能差异

能力 gpt-image-2 / gpt-image-2-pro 米醋 Grok 图像模型
默认用途 主通道,覆盖文生图、图生图、批量编辑、多图参考 可选通道,适合快速文生图、单图参考、多图参考
可选模型 gpt-image-2, gpt-image-2-pro grok-imagine-image-lite, grok-imagine-image, grok-imagine-image-pro, grok-imagine-image-edit
image_generate 文生图 支持 1K / 2K / 4K;2K/4K 自动切 pro,强制 n=1 支持 1K / 2K 路由;n 会传给后端,实际返回张数以响应为准
image_edit 单图参考/编辑 1K 走 /v1/images/edits;2K 走 reference_image;4K 参考图入口拒绝 /v1/images/generations + reference_image;4K 会映射到 resolution=2k
局部 mask 仅 1K edits multipart 支持 alpha mask;2K 不支持 当前不支持 mask,传入会忽略并写入 notes
image_multi_reference 多图参考 2-10 张参考图;1K 稳定,2K 可能 fallback,4K 入口拒绝 2-10 张参考图走 image_urls;实测可用,按 resolution + aspect_ratio 映射
image_batch_edit 批量逐张编辑 支持 1K;non-pro 5 并发,pro 串行 当前不支持 Grok 批量逐张编辑
size 校验 WxH,边长 256-4096,W/H 必须是 8 的倍数 只校验 WxH 正整数,不强制 8 倍数和 4096 边长
实际输出尺寸 ≥4MP 通常严格 1:1;≤2.25MP 会被代理处理到约 1.57MP 不保证等于请求 WxH,以 saved.actual_size 为准
重试/限流 2K/4K 使用跨进程锁,避免多个 MCP 同时打 pro 队列 不走高分辨率锁;可恢复错误仍自动重试并记录到 notes
配置变量 MICU_API_KEY, MICU_MODEL, MICU_BASEURL MICU_GROK_API_KEY, XAI_MODEL;默认复用 MICU_BASEURL

一键安装

git clone https://github.com/Subaru486desuwa/micu-image-mcp.git
cd micu-image-mcp
python install.py

脚本会:

  1. 检查 Python >= 3.10
  2. 安装依赖
  3. 交互配置米醋 API key、输出目录
  4. 可选配置米醋 Grok 生图 token
  5. 写入 ~/.claude.json~/.codex/config.toml
  6. 启动 server 做一次 initialize 握手

非交互安装:

MICU_API_KEY=sk-... \
MICU_GROK_API_KEY=sk-... \
MICU_SAVE_DIR=~/Pictures/micu-out \
python install.py --yes

常用选项:

python install.py --no-codex
python install.py --no-claude
python install.py --mirror tsinghua
python install.py --baseurl https://www.micuapi.ai

安装完成后重启 Claude Code / Codex,让 LLM 调 server_info 验证。


Grok 路径

Grok 走米醋中转,base URL 默认仍是:

https://www.micuapi.ai

只需要额外配置:

MICU_GROK_API_KEY=sk-...
XAI_MODEL=grok-imagine-image-lite
MICU_GROK_SIZE_MODE=contain

Grok 的 size 不套用 image2 的 8 倍数和 4096 边长约束。本地只检查 WxH 格式,然后映射为:

  • resolution: 1k2k
  • aspect_ratio: 最接近的比例,如 1:116:99:16

注意:Grok 后端返回像素不保证严格等于请求的 WxH。MCP 默认会在保存前用 Pillow 把 Grok 输出归一化到请求尺寸,MICU_GROK_SIZE_MODE 可选:

行为
contain 默认。等比缩放,补边到请求尺寸,不裁主体
cover 等比缩放并居中裁切,铺满请求尺寸
stretch 直接拉伸到请求尺寸,可能变形
backend 不做本地后处理,保留 Grok 后端原始像素

建议仍优先用常见比例和不太小的边长,例如 1024x10241536x10241024x15361501x1001。过小或很奇异的比例可能被米醋 Grok 后端返回 500,MCP 会自动重试并在 notes 里记录。


Size 规则

image2 路径:

  • W/H 必须是 8 的倍数
  • W/H 必须在 256 到 4096 范围内
  • 1K 福利档可能被代理处理到约 1.57MP
  • 2K/4K 自动切 gpt-image-2-pro
  • 2K/4K 强制 n=1 并加跨进程锁,避免多个 MCP 同时打爆 pro 队列

推荐 size:

档位 推荐值
1K 1024x1024, 1280x720, 720x1280, 1024x1536, 1536x1024
2K 2048x2048, 2048x1152, 1152x2048
4K 3840x2160, 2160x3840

Grok 路径:

  • 不强制 8 倍数
  • 当前按 1K / 2K 路由
  • 4K 请求会映射到 resolution=2k

环境变量

变量 默认值 说明
MICU_API_KEY 米醋 image2 token
MICU_BASEURL https://www.micuapi.ai 米醋 base URL
MICU_MODEL gpt-image-2 image2 默认模型
MICU_GROK_API_KEY 米醋 Grok 图像 token
XAI_MODEL grok-imagine-image-lite Grok 默认模型
MICU_GROK_SIZE_MODE contain Grok 保存前尺寸归一化策略:contain / cover / stretch / backend
MICU_SAVE_DIR ~/Pictures/micu-out 默认输出目录
MICU_SAVE_DIR_ROOT 同输出目录 输出安全根目录
MICU_USE_SHELL_PROXY 0 设为 1 才读取 shell 代理

兼容旧 Grok 变量 XAI_API_KEY / GROK_API_KEY,但推荐新配置统一使用 MICU_GROK_API_KEY


手动配置

Claude Code:

{
  "mcpServers": {
    "micu-image": {
      "command": "/path/to/python",
      "args": ["/absolute/path/to/micu-image-mcp/server.py"],
      "env": {
        "MICU_API_KEY": "sk-...",
        "MICU_GROK_API_KEY": "sk-...",
        "MICU_SAVE_DIR": "/Users/you/Pictures/micu-out",
        "MICU_SAVE_DIR_ROOT": "/Users/you/Pictures/micu-out",
        "XAI_MODEL": "grok-imagine-image-lite"
      }
    }
  }
}

Codex:

[mcp_servers.micu-image]
command = "/path/to/python"
args = ["/absolute/path/to/micu-image-mcp/server.py"]

[mcp_servers.micu-image.env]
MICU_API_KEY = "sk-..."
MICU_GROK_API_KEY = "sk-..."
MICU_SAVE_DIR = "/Users/you/Pictures/micu-out"
MICU_SAVE_DIR_ROOT = "/Users/you/Pictures/micu-out"
XAI_MODEL = "grok-imagine-image-lite"

Recommended Servers

playwright-mcp

playwright-mcp

A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.

Official
Featured
TypeScript
Magic Component Platform (MCP)

Magic Component Platform (MCP)

An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.

Official
Featured
Local
TypeScript
Audiense Insights MCP Server

Audiense Insights MCP Server

Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.

Official
Featured
Local
TypeScript
VeyraX MCP

VeyraX MCP

Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.

Official
Featured
Local
graphlit-mcp-server

graphlit-mcp-server

The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.

Official
Featured
TypeScript
Kagi MCP Server

Kagi MCP Server

An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.

Official
Featured
Python
E2B

E2B

Using MCP to run code via e2b.

Official
Featured
Neon Database

Neon Database

MCP server for interacting with Neon Management API and databases

Official
Featured
Exa Search

Exa Search

A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.

Official
Featured
Qdrant Server

Qdrant Server

This repository is an example of how to create a MCP server for Qdrant, a vector search engine.

Official
Featured