deepseek-vision-mcp
Click on "Deploy 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., "@deepseek-vision-mcpdescribe this image: /tmp/screenshot.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.
DeepSeek Vision MCP Server
将 DeepSeek 网页版识图(Vision)模式封装为 MCP 工具,解决 DeepSeek 模型无法直接处理图像的问题。
安装
git clone <repo> && cd deepseek-vision-mcp
npm install && npm run buildRelated MCP server: mcp-vision
使用
方式一:自动登录(推荐,桌面环境)
无需任何配置,直接启动:
npm start首次启动时自动打开浏览器让您登录 DeepSeek,Token 自动保存到 ~/.deepseek-vision/config.json,后续自动复用。
Token 过期后会自动弹出浏览器重新登录。
方式二:手动配置(服务器环境)
在 .mcp.json 中设置环境变量:
{
"mcpServers": {
"deepseek-vision": {
"command": "node",
"args": ["/path/to/dist/index.js"],
"env": {
"DEEPSEEK_USER_TOKEN": "你的token",
"DEEPSEEK_SMIDV2": "你的smidV2(可选)"
}
}
}
}Token 获取方法:https://chat.deepseek.com → F12 → Application → Local Storage → userToken → JSON.parse(value).value
MCP 工具
recognize_image
参数 | 类型 | 必填 | 说明 |
| string | ✅ | 绝对路径或 data URI。DeepSeek 模型识图必须走本工具;先 |
| string | ❌ | 提问(默认:请详细描述这张图片中的内容) |
模型规则: 当前对话使用 DeepSeek 模型(deepseek-chat、deepseek-reasoner 等)时,分析图片内容必须优先调用 recognize_image,不要让 DeepSeek 直接"看"图。
recognize_image({ image: "/tmp/screenshot.png", prompt: "这张图是什么" })项目结构
src/
├── index.ts # MCP Server 入口 + Skill 自动安装
├── deepseek-client.ts # API 客户端(上传/Fork/PoW/Completion)
├── auth-manager.ts # Puppeteer 自动登录 + Token 管理
├── pow-solver.ts # WASM PoW 求解器
├── types.ts # 类型定义
└── wasm/ # PoW 模块
SKILL.md # Agent Skill 定义启动时自动将 SKILL.md 安装到 ~/.agents/skills/deepseek-vision/SKILL.md。
This server cannot be deployed
Maintenance
Related MCP Connectors
Provides cloud browser automation capabilities using Stagehand and Browserbase, enabling LLMs to i…
Web search, browser automation, scraping, crawling and CAPTCHA solving for AI agents.
Web search and page-reading for AI agents. One-click OAuth connect, or a Caesar API key.
Stealth web automation for AI agents. Login, signup, navigate, screenshot.
Related MCP Servers
- FlicenseAqualityCmaintenanceEnables LLMs like DeepSeek to understand images by calling external vision models via OpenAI-compatible API. Provides tools to describe images or diagnose connectivity.2-
- FlicenseBqualityCmaintenanceEnables image recognition and description using Kimi K2.5 (Infini-AI) via the recognize_image tool. Supports local image files with optional custom prompts.1-
- FlicenseNot gradedqualityCmaintenanceEnables image recognition using vision models via OpenAI-compatible APIs, supporting multiple platforms like OpenAI, DeepSeek, and Ollama.-
- AlicenseAqualityBmaintenanceProvides image recognition for text-only LLMs like DeepSeek by bridging to SenseNova multimodal model, enabling image description via the describe_image tool.110 npm3MIT