Vision MCP Server
Provides image understanding capabilities using OpenAI's vision models (e.g., gpt-4o-mini) to analyze image content and return descriptions for software development contexts.
Click on "Install Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@Vision MCP Serverdescribe the image at ./ui_mockup.png"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
Vision MCP Server
为什么需要它
许多 AI 模型(包括部分 Claude 版本)不能直接"看"图片。当你给它一张 UI 截图或设计稿时,它无法理解画面内容,也就无法帮你分析界面、提取文字、写代码复刻。
Vision MCP Server 填补了这个缺口——它充当 AI 的"眼睛",让你在对话中直接丢图片给模型分析,同时提供缓存、模型回退、UI 对比等实用能力。
Related MCP server: vision-bridge-mcp
它能做什么
场景 | 用法 |
给我描述这张截图的内容 |
|
这张设计稿用了什么配色和字体 |
|
实现截图和设计稿对比,找出差异 |
|
额度用完了自动换模型接着跑 |
|
同一张图昨天分析过,今天免费用 | 磁盘缓存,7 天 TTL |
典型工作流
"生成这个页面的 React 代码"
↓ Claude 写完
↓ Playwright 截图
↓ image_compare(设计稿, 截图)
↓ 返回差异 → Claude 修复第一次分析 5-15 秒,第二次相同图 0.01 秒(命中原缓存),不花二次钱。
快速开始
pip install -e .
python -m vision_mcp_serverClaude Code 在项目根目录配 .mcp.json 即可用,不需要额外配置(前提是已设置 API Key)。
工具
image_understand — 单图分析
image_understand(image_path, prompt=None, mode="quick", force_refresh=False)参数 | 说明 |
| 本地图片路径或 HTTP URL |
| 自定义提问,不传则自动选择提示词 |
|
|
| 跳过缓存 |
image_compare — 双图对比
image_compare(expected_image, actual_image, mode="ui", force_refresh=False)返回结构化 JSON,包含差异类型、严重度、区域和预期/实际对比。
配置
环境变量
# 必选
DASHSCOPE_API_KEY=sk-xxx # 百炼 API Key(默认 Provider)
# Provider 切换
VISION_PROVIDER=bailian # bailian | openai | openrouter
VISION_API_KEY=sk-xxx # 覆盖 API Key
VISION_BASE_URL=https://... # 覆盖 API 端点
# 模型回退
VISION_MODELS=qwen-vl-max,qwen-vl-plus,qwen-3.7 # 按顺序尝试,遇到 quota/429 自动切换
# 缓存
VISION_CACHE_ENABLED=false # 禁用缓存
VISION_CACHE_TTL=604800 # TTL(秒),默认 7 天
VISION_CACHE_DIR=/path/to/cache # 自定义缓存目录Provider 默认值
Provider | 模型 | API 地址 |
|
|
|
|
|
|
|
|
|
扩展
启动 HTTP 模式调试:
python -m vision_mcp_server --httpThis server cannot be installed
Maintenance
Resources
Unclaimed servers have limited discoverability.
Looking for Admin?
If you are the server author, to access and configure the admin panel.
Latest Blog Posts
- Your AI Chatbot Just Exposed Your CEO's Salary to an InternBy Om-Shree-0709 on .Agent IdentityMCP SecurityOAuth Delegation
- Why MCP Servers Need Execution Sandboxing (And Why Your Current Stack Isn't Enough)By Om-Shree-0709 on .Agentic AiPrompt InjectionWebAssembly
MCP directory API
We provide all the information about MCP servers via our MCP API.
curl -X GET 'https://glama.ai/api/mcp/v1/servers/GuanJiu173/DeepSeek-Vision-MCP'
If you have feedback or need assistance with the MCP directory API, please join our Discord server