Safe Vision MCP
Allows using OpenAI-compatible vision models to generate descriptions of images, with configurable API endpoint, model, and API key.
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., "@Safe Vision MCPReview images in my screenshots folder"
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.
Safe Vision MCP
通过 MCP 安全、高效地查看本地图片,避免把大图片直接塞进主模型上下文。
Safe Vision MCP 可以生成缩略图、联系表、图片统计信息、清单文件,以及可选的视觉模型描述。工具返回的结果只包含路径和简短文本,不会返回图片字节、base64 或 data URL。
为什么需要它?
大截图和批量图片很容易占满 LLM 的上下文窗口。Safe Vision MCP 通过以下方式保持上下文轻量:
将派生文件写入本地磁盘;
只返回文件路径、数量、统计信息和简短描述;
使用缩略图和联系表快速浏览图片;
在 MCP 服务器内部,每次最多只向视觉模型发送一张原图或派生图。
Related MCP server: image-tiler-mcp-server
工具列表
工具 | 用途 |
| 对指定图片路径或目录进行一站式批量审查 |
| 查看单张图片的尺寸、基础统计信息和缩略图 |
| 生成缩略图、联系表、特征 CSV 和 manifest |
| 为指定图片生成联系表 |
| 从目录中确定性地抽取代表性图片 |
环境要求
Node.js 18+
Python 3.10+
Python 包:
Pillow
npm install
python -m pip install -r requirements.txtMCP 配置示例
[mcp_servers.safe_vision]
command = "node"
args = ["/absolute/path/to/safe-vision-mcp/src/server.js"]
cwd = "/absolute/path/to/safe-vision-mcp"
type = "stdio"
startup_timeout_sec = 20
tool_timeout_sec = 3600
[mcp_servers.safe_vision.env]
SAFE_VISION_PYTHON = "python"
SAFE_VISION_OPENAI_BASE_URL = "https://api.openai.com/v1"
SAFE_VISION_MODEL = "gpt-5.4-mini"
SAFE_VISION_OPENAI_API_KEY_ENV = "OPENAI_API_KEY"如需使用 OpenAI 兼容服务,请配置:
SAFE_VISION_OPENAI_BASE_URLSAFE_VISION_MODELSAFE_VISION_OPENAI_API_KEY或SAFE_VISION_OPENAI_API_KEY_ENV
不要将私有接口地址、模型别名、.env 文件或 API Key 提交到公开仓库。
环境变量
名称 | 默认值 | 作用 |
|
| Node MCP 服务器调用的 Python 启动器 |
| 空 | 视觉模型服务的直接 API Key |
| 空 | 指向 API Key 的环境变量名 |
| 空 | 备用 API Key |
|
| OpenAI 兼容 Responses API 地址 |
| 空 | 备用 API 地址 |
|
| MCP 服务器内部使用的视觉模型 |
|
| 单张图片的视觉模型请求超时 |
|
|
|
|
|
|
视觉输入策略
启用视觉模型描述时,MCP 服务器会向配置的视觉服务发送以下图片之一:
原图:当图片格式为 JPEG/PNG/WebP、长边
<= 2048px、文件大小<= 8MB;派生 JPEG:保存在
vision_inputs/下,长边缩放到2048px,应用 EXIF 方向,质量为90。
无论哪种模式,MCP 返回结果仍然只包含路径、文本、数量和紧凑统计信息。
输出内容
review_images 会生成:
descriptions.json
manifest.json
features.csv
thumbnails/
sheets/
vision_inputs/ # 可选每条图片描述包含:
indexpathfilenamethumbnail_pathdescription_sourcedescriptionfeatures
语义输出统一放在单个 description 字符串中,没有单独的 observation 对象。
隐私说明
MCP 响应永远不会包含图片字节、base64 字符串或 data URL。
视觉模式只会把图片内容发送给你在私有环境变量中配置的服务商。
启发式模式不会调用远程模型,只在本地生成派生文件。
请勿提交
.env、私有 MCP 配置、私有 base URL、模型别名或 API Key。
快速测试
node --check src/server.js
python -m py_compile src/vision_tools.py如需在不调用远程模型的情况下测试,可以在 MCP 客户端中调用 review_images,并设置:
{
"describe_mode": "heuristic"
}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 Servers
- AlicenseAqualityDmaintenanceMCP OpenVision is a Model Context Protocol (MCP) server that provides image analysis capabilities powered by OpenRouter vision models. It enables AI assistants to analyze images via a simple interface within the MCP ecosystem.Last updated115MIT
- AlicenseAqualityCmaintenanceMCP server that gives LLMs full-resolution vision by tiling images and capturing web pages before details are lost.Last updated1392MIT
- AlicenseAqualityDmaintenanceAn MCP server that provides classic computer vision feature extraction tools (color, texture, shape) for image analysis, enabling LLMs to perform precise mathematical image comparisons and quality assessments.Last updated9MIT
- Flicense-qualityCmaintenanceA low-hallucination vision MCP server that uses OpenAI-compatible multimodal models with structured prompts, confidence gating, and forced JSON to reduce false claims in image analysis.Last updated
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.
Remote MCP for C2PA intake verifier MCP, structured receipts, audit logs, and reviewer-ready evidenc
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/williamxhero/SafeVision'
If you have feedback or need assistance with the MCP directory API, please join our Discord server