dsh-vision-mcp
Allows AI agents to analyze images, extract text, and compare images using an OpenAI-compatible vision API endpoint.
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., "@dsh-vision-mcpextract all text from the image at C:/shots/photo.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.
dsh-vision-mcp
本地识图 MCP 服务器:给任何 MCP 客户端的 agent 装上眼睛——DeepSeek Harness、Claude Desktop、Codex CLI、Cursor 通用。视觉大脑是你自己的 OpenAI 兼容视觉端点(默认对接本机 chatgpt2api,也可指向任何标准 vision API),主模型无需自带视觉能力。
任意 MCP 客户端(stdio JSON-RPC)
│
┌───────▼────────────────────┐ HTTP POST /v1/chat/completions
│ dsh-vision-mcp │ ──────────────────────────────→ OpenAI 兼容视觉端点
│ analyze_image/extract_text/│ {image_url: dataURL} (chatgpt2api :8000
│ compare_images │ 或任何标准 vision API)
└────────────────────────────┘快速开始
git clone https://github.com/alwaysalone1/dsh-vision-mcp.git
cd dsh-vision-mcp
npm install && npm run build
# 环境体检:端点可达性 + key 校验 + 模型列表
export VISION_API_KEY='<你的 chatgpt2api auth-key>'
npm run doctorRelated MCP server: vision-bridge-mcp
接入各客户端
以下配置通用:command 跑构建产物,env 传入视觉后端的地址与 key。
DeepSeek Harness(packages/mcp 桥,一行 patch,工具名为 mcp__vision__analyze_image 等):
- insert:
- id: vision-mcp
name: '@deepseek-ai/dsh-mcp-client'
config:
serverName: vision
transport: stdio
command: node
args: ['D:/path/to/dsh-vision-mcp/dist/index.js']
env:
VISION_API_BASE: http://127.0.0.1:8000
VISION_API_KEY: !!js process.env.VISION_API_KEYClaude Desktop / 通用 JSON(claude_desktop_config.json 的 mcpServers,Cursor 的 mcp.json 同构):
{
"mcpServers": {
"dsh-vision": {
"command": "node",
"args": ["D:/path/to/dsh-vision-mcp/dist/index.js"],
"env": {
"VISION_API_BASE": "http://127.0.0.1:8000",
"VISION_API_KEY": "<your-key>"
}
}
}
}Codex CLI(~/.codex/config.toml):
[mcp_servers.dsh-vision]
command = "node"
args = ["D:/path/to/dsh-vision-mcp/dist/index.js"]
env = { "VISION_API_BASE" = "http://127.0.0.1:8000", "VISION_API_KEY" = "your-key" }工具
工具 | 作用 | 参数 |
| 理解图片:描述内容、回答问题、定位界面元素(可问"红色按钮的坐标") |
|
| OCR:按原始排版提取全部文字(代码/公式/多语言) |
|
| 对比两张图片的相同与不同 |
|
source 四种写法任选:本地文件路径(如 C:/shots/1.png)、http(s):// 图片 URL、data:image/png;base64,...、裸 base64。格式按文件头魔数判定(PNG/JPEG/GIF/WebP/BMP),不看扩展名。
模型选择
默认使用 gpt-5-6(账号池里最强的推理+视觉对话模型)。三种调整方式:
每次调用传
model参数临时覆盖(如gpt-5-6-mini省额度);全局默认:环境变量
VISION_MODEL(如设为auto交由后端路由);注意
gpt-image-2是生图模型,不用于识图。
配置(环境变量)
变量 | 默认 | 说明 |
| 必填 | 视觉后端的 Bearer key(chatgpt2api 的 |
|
| OpenAI 兼容端点地址 |
|
| 全局默认模型 |
|
| 单次请求硬超时(账号池繁忙时可调大) |
|
| 单图字节上限(编码前) |
|
| 传输错误/429/5xx 的重试次数(退避递增) |
排障
token_expired—— 账号池上游 token 过期:到 chatgpt2api 面板刷新/重登账号(配置里有auto_relogin_after_refresh)。HTTP 502 / TLS connect error —— 池子的上游代理(WARP/Privoxy)不健康:检查 chatgpt2api 的稳定代理运行时面板。
HTTP 401 ——
VISION_API_KEY与后端auth-key不一致。请求超时 —— 调大
VISION_TIMEOUT_MS;池子空闲时通常 10-30 秒内返回。格式不识别 —— 图片按文件头判定;损坏或非图片文件会被拒绝。
隐私与边界
发给视觉端点的图片会到达其后端模型(默认为你自己的 chatgpt2api 账号池→ChatGPT):不要用公共 vision API 处理敏感截图。
本服务器只做工具桥接:无状态、不落盘、不写日志文件(诊断仅到 stderr)。
开发
npm run typecheck # 严格模式类型检查
npm run build # 产物到 dist/
npm run smoke # 真实 stdio 端到端冒烟(需 VISION_API_KEY + 测试图)
npm run doctor # 环境体检许可
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections. Dates show when Glama detected each change.
No tool schema history has been recorded yet.
This 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.
Related MCP Connectors
OCR, transcription, file extraction, and image generation for AI agents via MCP.
Use AI models for chat, image, and video generation from Claude Code and other MCP hosts.
Focused MCP server for OpenAI image/audio generation (v2.0.0). Wraps endpoints via HAPI CLI.
Connect MCP clients to 2,000+ AI models without managing provider API keys.
Related MCP Servers
- AlicenseAqualityCmaintenanceGive MCP-compatible AI agents image analysis, metadata inspection, cropping, OCR, and image comparison through any OpenAI-compatible vision model.6MIT
- AlicenseAqualityBmaintenanceEnables non-multimodal models to see images by providing MCP tools for image understanding and OCR, backed by any OpenAI-compatible vision model.2MIT
- AlicenseAqualityBmaintenanceEnables any MCP client to perform image understanding and OCR via any OpenAI-compatible vision-language model. Supports local, private inference without images leaving the machine.232MIT
- AlicenseAqualityCmaintenanceProvides image recognition capabilities to MCP clients by integrating with OpenAI-compatible vision models, supporting local images, URLs, multi-image comparison, and model listing.4MIT
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/alwaysalone1/dsh-vision-mcp'
If you have feedback or need assistance with the MCP directory API, please join our Discord server