llm-vision-mcp
Allows the MCP server to use a local Ollama vision model as the image analysis backend, enabling text-only LLMs to describe images through Ollama's OpenAI-compatible API.
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., "@llm-vision-mcpDescribe the image at /home/user/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.
llm-vision-mcp 👁️
📌 这是什么?
通过 MCP stdio 协议为文本模型提供图片理解能力的本地服务,专注磁盘图片分析:
支持任意 OpenAI 兼容的视觉模型供应商(默认 ModelScope 免费的通义千问VL,国内直连,每天 2000 次调用,单模型 500 次),可通过
BASE_URL/VISION_MODEL切换通过 MCP stdio 协议供 Claude Code / Opencode 等客户端调用
本地进程运行,不开放任何网络端口
为什么需要? DeepSeek / GLM 等文本模型的 API 没有视觉能力——给它一个图片路径,它只能"看到"路径本身。本 MCP Server 把图片转成文字描述,让文本模型也能"看图说话"。
DeepSeek / Claude Code / Opencode
│ MCP stdio
▼
┌─────────────────────┐ HTTPS ┌──────────────────────┐
│ llm-vision-mcp │ ────────▶ │ OpenAI 兼容供应商 │
│ (Node.js) │ │ (默认 ModelScope │
└─────────────────────┘ │ Qwen3-VL-8B) │
▲ └──────────────────────┘
│ 读取图片文件 → base64 → 返回文字描述Related MCP server: Vision MCP Server
⚡ 快速开始
已发布到 npm,无需克隆仓库,客户端会通过 npx / pnpm dlx 自动下载运行:
# 首次使用:获取免费 API Key(每天2000次,单模型500次)—— 默认 ModelScope 供应商
# ① 打开 https://modelscope.cn 注册/登录
# ② 点右上角头像 → 个人中心 → 访问令牌
# 或直接访问: https://modelscope.cn/my/myaccesstoken
# ③ 首次使用会提示绑定阿里云账号(必须,按页面引导完成)
# ④ 点击"新建访问令牌" → 命名 → 生成 → 复制
# ⑤ 令牌格式为 ms-xxxxxxxxxxxx,使用时去掉 ms- 前缀!
# 验证可用性:启动 MCP Server(stdio,Ctrl+C 退出)
npx -y @me9rez/llm-vision-mcp mcp
# 或 pnpm
pnpm dlx @me9rez/llm-vision-mcp mcp然后按下方「客户端配置」把 MCP Server 接入 Claude Code / Opencode,在 env 中注入 API_KEY 即可。也可以直接用 CLI 分析图片(见下方「🖥️ CLI 使用」)。
🧑💻 本地开发
git clone <仓库地址> && cd llm-vision-mcp
npm install
npm start # 启动 MCP Server
npm test # 单元测试
npm run smoke # 冒烟测试(无需 API Key)🖥️ CLI 使用
不接 MCP 客户端时,也可以在终端直接分析图片:
# 查看帮助(列出全部命令与环境变量)
npx -y @me9rez/llm-vision-mcp --help
# 分析图片(默认通用分析提示词)
npx -y @me9rez/llm-vision-mcp analyze_image ./图片.png
# 带自定义问题
npx -y @me9rez/llm-vision-mcp analyze_image ./图片.png "描述图片中的颜色和字体"
# 其他命令:extract_text(OCR)、describe_ui、diagnose_error、
# understand_diagram、analyze_chart、code_from_screenshot
# 下划线可替换为短横线(如 extract-text),效果相同
# 启动 MCP Server(stdio 模式)
npx -y @me9rez/llm-vision-mcp mcpAPI_KEY=你的密钥 npx -y @me9rez/llm-vision-mcp analyze_image ./图片.png📋 MCP 工具列表
工具 | 功能 |
| 分析磁盘图片文件(可传自定义 |
| 磁盘图片 OCR 提取文字 |
| 分析磁盘 UI 图片 |
| 诊断磁盘错误图片 |
| 解读流程图/架构图 |
| 分析数据图表 |
| 从磁盘图片提取代码 |
🔌 客户端配置
Claude Code(.claude/settings.json):
{
"mcpServers": {
"llm-vision-mcp": {
"command": "npx",
"args": ["-y", "@me9rez/llm-vision-mcp", "mcp"],
"env": {
"API_KEY": "你的_API_Key"
}
}
}
}⚠️ Windows 下如
npx无法直接启动,可将command改为npx.cmd,或使用"command": "cmd", "args": ["/c", "npx", "-y", "@me9rez/llm-vision-mcp", "mcp"]。 完整示例见examples/claude_code_settings.json。
只启用部分工具(减少 agent 上下文占用,TOOLS 与 --tools 二选一):
{
"mcpServers": {
"llm-vision-mcp": {
"command": "npx",
"args": ["-y", "@me9rez/llm-vision-mcp", "mcp", "--tools", "analyze_image,extract_text"],
"env": {
"API_KEY": "你的_API_Key"
}
}
}
}或通过 env 注入:"env": { "API_KEY": "…", "TOOLS": "analyze_image,extract_text" }。工具名支持 - 代替 _(如 extract-text)。
Opencode(%APPDATA%\opencode\opencode.json):
{
"mcp": {
"llm-vision-mcp": {
"type": "local",
"command": ["npx", "-y", "@me9rez/llm-vision-mcp", "mcp"],
"enabled": true,
"environment": {
"API_KEY": "你的_API_Key"
}
}
}
}完整示例见
examples/opencode.json。
⚙️ 环境变量
支持任意 OpenAI 兼容的视觉模型供应商,全部通过环境变量配置:
变量 | 必填 | 默认值 | 说明 |
| 是 | - | 供应商密钥(ModelScope 令牌需去掉 |
| 否 |
| OpenAI 兼容接口地址,可替换为任意供应商 |
| 否 |
| 视觉模型名(如 |
| 否 |
| 采样温度(数字,如 |
| 否 |
| 最大生成长度(正整数) |
| 否 | (全部) | 启用工具白名单,逗号分隔(如 |
配置其他供应商示例(如本地 Ollama / vLLM 部署):
{
"mcpServers": {
"llm-vision-mcp": {
"command": "npx",
"args": ["-y", "@me9rez/llm-vision-mcp", "mcp"],
"env": {
"API_KEY": "sk-xxx",
"BASE_URL": "http://localhost:8000/v1",
"VISION_MODEL": "qwen2.5-vl-7b",
"TEMPERATURE": "0.2",
"MAX_TOKENS": "32768"
}
}
}
}🛡️ 安全
本地 stdio 进程运行,不开放任何网络端口
仅接受图片格式(
.png .jpg .jpeg .gif .webp .bmp),防止 LLM 注入后读取任意文件文件大小限制 20MB,扩展名 + 魔数双重校验
图片经 base64 编码发送至视觉模型供应商 API,参阅其隐私政策
🙏 参考来源
deepseek-eyes(MIT)— 本项目参考的原始项目:MCP Server + 通义千问VL,原项目为 Python 实现并含剪贴板工具;本项目以 Node.js 重写并移除了剪贴板工具
视觉模型:通义千问VL / Qwen-VL via ModelScope
📄 License
MIT
Maintenance
Resources
Unclaimed servers have limited discoverability.
Looking for Admin?
If you are the server author, to access and configure the admin panel.
Related MCP Servers
- AlicenseAqualityBmaintenanceBridges a vision model to enable text-only models like DeepSeek to describe images, extract text, and compare images via MCP tools.Last updated5678MIT
- Flicense-qualityBmaintenanceProvides image understanding capabilities for MCP clients (e.g., Claude Code) by analyzing images using vision models from providers like Alibaba Cloud Bailian, OpenAI, or OpenRouter, returning detailed descriptions in Markdown format.Last updated1
- FlicenseAqualityBmaintenanceGives any MCP client (OpenCode, Claude Code, Claude Desktop, Cursor, etc.) the ability to process images by automatically converting them to text descriptions using a vision model, so that text-only LLMs can handle image-based queries.Last updated2
- AlicenseAqualityBmaintenanceBridges vision models to text-only coding models using Florence-2, enabling non-vision LLMs to describe images, extract text, and analyze screenshots via MCP tools.Last updated6MIT
Related MCP Connectors
OCR, transcription, file extraction, and image generation for AI agents via MCP.
Zero-auth MCP: image optimize, cited storage/format data, and dev utilities LLMs get wrong.
Multimodal video analysis MCP — transcription, vision, and OCR for any video URL.
Latest Blog Posts
- Who's Calling? MCP Hosts Are an Identity Blind Spot (And the Spec Knows It)By Om-Shree-0709 on .mcpAgent IdentityOAuth 2.1
- 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/me9rez/llm-vision-mcp'
If you have feedback or need assistance with the MCP directory API, please join our Discord server