mimo-vision
Provides Claude Code with visual capabilities by analyzing local images via Xiaomi MiMo-V2.5, enabling image analysis, comparison, and UI inspection to guide frontend development.
README
mimo-vision MCP Server
本地 stdio MCP Server:给 Claude Code(主模型可以是 DeepSeek 等任何模型)提供"视觉眼睛"。 把本地图片交给小米 MiMo-V2.5 图片理解模型分析,返回文本结果供主 Agent 继续工作。
- 安装位置:
C:\Users\MR\.claude\mcp\mimo-vision - 视觉模型:
mimo-v2.5(官方 OpenAI-compatible 接口https://api.xiaomimimo.com/v1/chat/completions,直接调用无中间层) - 安全:只做"读图片 → 调 MiMo API → 返回文本",无文件写入/删除/shell 能力
- API Key 运行时从环境变量
MIMO_API_KEY读取,绝不写入源码、配置或日志
配置 MIMO_API_KEY(Windows)
PowerShell 或 CMD 执行:
setx MIMO_API_KEY "你的Key"
然后完全退出并重启 Claude Code(MCP 无需重新注册)。
注册到 Claude Code(user scope,所有项目可用)
claude mcp add -s user mimo-vision -- "C:\Program Files\nodejs\node.exe" "C:\Users\MR\.claude\mcp\mimo-vision\server.js"
验证:
claude mcp list # 应显示: mimo-vision ✔ Connected
三个 Tool
| Tool | 参数 | 用途 |
|---|---|---|
analyze_image |
image_path 必填;prompt 可选 |
分析一张图:布局、排版、间距、配色、视觉问题 |
compare_images |
image_paths 必填(2~8 张,按顺序);prompt 可选 |
一次跨图对比:相同点/差异/布局/排版/间距/颜色/组件/比例/层级 + 对改代码最有价值的信息 |
inspect_ui |
image_path 或 image_paths;task 可选;project_context 可选 |
UI 工程化分析,13 项结构化结果(Layout、尺寸比例、Grid、Spacing、Typography、Colors、边框圆角阴影、组件层级、图片比例、响应式线索、交互线索、差异问题、实现建议),可直接用于写 HTML/CSS/React |
图片路径支持:Windows 绝对路径(含中文/空格/反斜杠)或相对项目根目录的路径。 格式:JPEG/JPG/PNG/WebP/GIF/BMP;单张 Base64 不超过 50MB(超限返回明确错误并提示压缩)。
使用示例(对主 Agent 直接说)
- "看一下 ./screenshots/home.png,分析为什么这个页面显得太拥挤。"
- "对比 ./current.png 和 ./reference.png,然后修改当前页面使它更接近参考图。"
- "检查 ./result.png,看看这次前端修改还有哪些视觉问题。"
- "我要复刻 ./design.png 这个页面,先做 UI 分析再写代码。"
自测
cd C:\Users\MR\.claude\mcp\mimo-vision
node test/mcp-client-test.mjs # 14 项基础+失败测试(无需 Key)
node test/mcp-client-test.mjs --real # 再加真实 MiMo 调用(需已设置 MIMO_API_KEY)
测试图生成:node test/make-test-images.mjs(纯 Node 零依赖生成 PNG)。
排错
| 现象 | 处理 |
|---|---|
报错 MIMO_API_KEY 未设置 |
按上文 setx 设置后重启 Claude Code |
报错 MiMo 返回了空内容 |
常见于部分 WebP 编码(如带 ICC profile)MiMo 无法解析:把图片转成 PNG 后重试(可用图片查看器另存为,或 msedge --headless=new --disable-gpu --screenshot=输出.png --window-size=<宽>,<高> "file:///原图") |
报错 [HTTP 401] |
Key 错误或失效,到小米 MiMo 开放平台检查 |
报错 [HTTP 429] |
限流,稍后重试(服务端已自动重试 2 次) |
报错 图片过大 |
压缩/缩小图片(单张 Base64 ≤ 50MB) |
报错 不支持的图片格式 |
转成 PNG/JPG 等支持的格式 |
| 网络错误 | 检查网络/代理;服务端对网络错误自动退避重试 |
卸载
claude mcp remove mimo-vision
然后删除目录 C:\Users\MR\.claude\mcp\mimo-vision;如不再使用 MiMo,再执行 setx MIMO_API_KEY ""。
Recommended Servers
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.
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.
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.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
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.
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.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
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.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.