image-vision-mcp
Integrates with OpenAI-compatible vision models to generate detailed text descriptions of images from local files, URLs, or base64 data URIs.
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., "@image-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.
# image-vision-mcp
一个轻量的 stdio MCP server,给你的 AI 编程助手(Codex、Claude Code 等)加上看图能力。
它只暴露一个工具 describe_image:接收本地文件路径、HTTP(S) 链接或 base64 data URI,发给任意 OpenAI 兼容的 /v1/chat/completions 多模态端点,返回一段详细的文字描述,供调用方模型使用。
无需原生二进制、无需 Python —— 整个服务只有一个 Node.js 文件和唯一依赖 @modelcontextprotocol/sdk。
功能特性
describe_image工具,后端为 OpenAI 兼容的视觉模型支持 本地文件路径、http(s) 链接、base64 data URI 三种输入
内置 20 MB 图片体积上限保护
全部参数通过环境变量配置(模型、端点、鉴权 token)
兼容任何 MCP 客户端:Codex、Claude Code 等
Related MCP server: read-image-mcp
运行环境要求
Node.js
>= 18(需要内置全局fetch)一个 OpenAI 兼容的多模态端点。可以是:
本地代理,例如 CC Switch,运行在
127.0.0.1:15721云厂商(OpenAI 等)
自建网关
安装
git clone https://github.com/wxyhaha/image-vision-mcp.git
cd image-vision-mcp
npm install无需全局安装 —— MCP 客户端通过
node server.js启动本服务。
配置说明
全部通过环境变量配置:
变量名 | 默认值 | 说明 |
|
| OpenAI 兼容端点的根地址(本服务会自动拼接 |
|
| 要调用的多模态模型名称。 |
| (空) | 作为 |
在 Codex 中使用
把下面这段加到 ~/.codex/config.toml(Windows 下为 %USERPROFILE%\.codex\config.toml):
[mcp_servers.image_vision]
args = ['<本仓库的绝对路径>/server.js']
command = 'node'
startup_timeout_sec = 15
[mcp_servers.image_vision.env]
VISION_MODEL = "mimo-v2.5"
VISION_PROXY_BASE = "http://127.0.0.1:15721"
# 可选: VISION_PROXY_TOKEN = "sk-..."然后重启 Codex,给它发一张图片,模型会自动调用 describe_image。
直连端点(不走本地代理)
[mcp_servers.image_vision.env]
VISION_MODEL = "gpt-4o"
VISION_PROXY_BASE = "https://api.openai.com/v1"
VISION_PROXY_TOKEN = "sk-your-key-here"在 Claude Code 中使用
添加一个指向本仓库的 MCP server:
claude mcp add image-vision-mcp -- node /本仓库的绝对路径/image-vision-mcp/server.js然后设置环境变量(在你的 shell 或 Claude Code 配置里):
export VISION_PROXY_BASE="http://127.0.0.1:15721"
export VISION_MODEL="mimo-v2.5"
export VISION_PROXY_TOKEN="sk-..." # 可选验证安装是否成功
npm run check # 对 server.js 做语法检查
npm start # 应正常启动并在 stdio 上等待,按 Ctrl+C 退出完整链路冒烟测试:通过你的 MCP 客户端发一张图,看是否返回文字描述即可。
工作原理
MCP 客户端收到模型给出的图片引用。
调用本服务的
describe_image,传入image_path(和可选的question)。本服务读取文件(或拉取 URL),编码为 base64 data URI,以
VISION_MODEL指定的模型 POST 到${VISION_PROXY_BASE}/v1/chat/completions。返回的文字描述作为工具输出交还给调用方模型。
许可协议
MIT
This server cannot be deployed
Maintenance
Related MCP Connectors
Focused MCP server for OpenAI image/audio generation (v2.0.0). Wraps endpoints via HAPI CLI.
OCR, transcription, file extraction, and image generation for AI agents via MCP.
A comprehensive Model Context Protocol (MCP) server that enables AI assistants to interact with yo…
MCP server for Qwen Image 3 AI image generation
Related MCP Servers
- AlicenseBqualityCmaintenanceA lightweight MCP server for image analysis using any OpenAI-compatible API endpoint, enabling AI agents to analyze images via a single tool.119 npmMIT
- FlicenseBqualityDmaintenanceA stdio MCP server that enables agents without native vision to ask an OpenAI-compatible vision model to describe, OCR, or extract structured data from images via a single tool.1-
- AlicenseAqualityBmaintenanceProvides vision understanding capabilities such as image analysis, OCR, object localization, and video frame analysis, plus optional image generation and editing, to coding agents via OpenAI-compatible multimodal models. Runs as a local MCP server with HTTP and stdio transports, configurable for clients like Codex, Claude Code, Kimi, and Cursor.310 npmMIT
- AlicenseAqualityAmaintenanceMCP server that provides vision capabilities to coding agents, enabling them to analyze screenshots, UI mockups, terminal errors, documents, tables, and charts through OpenAI-compatible vision models. Supports local stdio and remote HTTP deployments with structured JSON output and binary upload side channels.820 npmMIT