vision-mcp
Allows converting images to text descriptions via an OpenAI-compatible vision model API, with configurable base URL, API key, and model.
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., "@vision-mcp Describe this image: /Users/me/Pictures/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.
vision-mcp
一个 MCP Server,把图片(JPEG / PNG / GIF / WebP)转换为文本描述。基于 OpenAI 兼容的视觉模型接口(默认 DeepSeek 的 deepseek-v4-flash-vision-exp)。
功能
describe_image— 输入图片(本地绝对路径或 http(s) URL),返回文本描述。configure— 保存 / 更新baseUrl/apiKey/model到.env文件。首次使用自动配置 — 首次启动时若
.env不存在,会自动生成模板并在日志中提示。
Related MCP server: image_mcp
安装与运行
npm install
npm run build
npx . # 在项目目录内运行(走 package.json 的 bin)
# 或
npm start # 等价于 node dist/index.js发布到 npm 后可直接:
npx vision-mcp配置
配置通过 .env 文件保存,读取优先级为:环境变量 > .env 文件 > 默认值(仅 model)。
变量 | 说明 | 必填 |
| OpenAI 兼容接口地址,默认 | 否 |
| API 密钥 | 是 |
| 视觉模型名,默认 | 否 |
三种配置方式:
调用
configure工具(推荐,在 MCP 客户端里调用即可):{ "apiKey": "sk-..." }编辑
.env文件:首次启动会自动在包目录生成模板,填入后重启即可。环境变量:直接设置
BASE_URL/API_KEY/MODEL(优先级最高)。
.env 的位置
默认保存在包自身目录下(由 import.meta.url 解析到包根目录)。注意:npx <pkg> 会把包装进 ~/.npm/_npx/<hash>/... 缓存,缓存清理后 .env 会丢失。
如需固定位置,设置环境变量 VISION_MCP_HOME,例如:
VISION_MCP_HOME="$HOME/.vision-mcp" npx vision-mcp此时 .env 会保存为 ~/.vision-mcp/.env,跨 npx 缓存清理依然保留。
工具说明
describe_image
参数 | 类型 | 说明 |
| string | 本地绝对路径,或 http(s) URL |
| string(可选) | 关于图片的提问/指令,默认 |
configure
参数 | 类型 | 说明 |
| string(可选) | 接口地址 |
| string(可选) | API 密钥 |
| string(可选) | 模型名 |
接入 Claude Desktop
在 claude_desktop_config.json 中添加:
{
"mcpServers": {
"vision-mcp": {
"command": "npx",
"args": ["vision-mcp"],
"env": {
"VISION_MCP_HOME": "/Users/你的用户名/.vision-mcp"
}
}
}
}本地开发调试时也可直接指向编译产物:
{
"mcpServers": {
"vision-mcp": {
"command": "node",
"args": ["/绝对路径/vision_mcp/dist/index.js"]
}
}
}开发
npm run dev # 开发模式(tsx 直跑源码)
npm run build # tsc 编译到 dist/
npm test # 运行单元测试(node:test + tsx)支持的图片格式
JPEG(jpg/jpeg)、PNG(png)、GIF(gif)、WebP(webp)。本地文件按扩展名识别;URL 优先使用响应的 content-type,否则回退到 URL 扩展名。
Available Tools
2 toolsconfigureA
Save or update the vision-mcp configuration (baseUrl, apiKey, model) to the .env file.
Inputs: model, apiKey, baseUrl.
describe_imageA
Convert an image (JPEG, PNG, GIF or WebP) into a text description using a vision model. The image may be a local absolute file path or an http(s) URL.
Inputs: image (required), prompt.
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
TDQS
Scored across 2 tools
The two tools are completely distinct: describe_image handles image-to-text conversion, while configure manages server settings. There is no overlap or ambiguity between them.
Both tool names are lowercase with underscores, and describe_image follows the verb_noun pattern. 'configure' is a simple verb without a noun, which is a minor deviation but still clear and predictable.
The server has only two tools, which feels slightly thin for a vision-oriented MCP. However, the narrow scope of 'describe an image' plus configuration is coherent, even if minimal.
The core operation (image description) is covered, and configuration is available, but obvious extensions like model listing or multiple vision capabilities (e.g., OCR) are missing. It works for basic use, but the surface is quite limited.
Maintenance
Related MCP Connectors
Generate images with your own ChatGPT subscription (Plus, Pro or Team), without spending API credits
Image & PDF tools for AI agents: compress, convert, resize, PDF, AI vision, pipeline.
Convert and resize static image URLs to WebP, PNG or JPEG for uploads and agent workflows.
Video, audio, and image processing for AI agents: convert, transcribe, upscale - 150+ operations.
Related MCP Servers
- AlicenseAqualityDmaintenanceEnables image analysis and understanding using Vision Language Models through OpenAI-compatible APIs. Supports analyzing images from URLs or local files with custom prompts.12MIT
- AlicenseNot gradedqualityDmaintenanceEnables text-only LLMs to analyze images by routing them to an OpenAI-compatible vision backend, supporting local files, URLs, and data URLs.13 npmMIT
- AlicenseAqualityDmaintenanceEnables image analysis using any OpenAI-compatible vision API, supporting URLs, local files, or base64 input with custom prompts.1MIT
- AlicenseNot gradedqualityDmaintenanceEnables image recognition using vision models via OpenAI-compatible APIs, supporting multiple platforms like OpenAI, DeepSeek, and Ollama.ISC