A
licenseA
qualityB
maintenanceGives text-only coding agents the ability to 'see' images, videos, and screenshots by routing them to a vision model and returning structured text.
8
32
1
MIT
visual-understand-mcp
by bcdxc
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., "@visual-understand-mcpWhat does this screenshot show?"
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.
视觉理解 MCP Server
为不支持视觉的编码模型提供图片理解能力。
原理
编码模型（如 GLM-5.1）遇到图片时，自动调用 MCP 工具 understand_image，由视觉模型完成图片识别，结果以文本回传给编码模型继续推理。无需切换模型，上下文不中断。
用户粘贴截图 → 编码模型调用 understand_image → 视觉模型识别 → 文字描述回传 → 继续编码Related MCP server: multimodal-mcp
安装方式一：通过 PyPI / uvx 使用（推荐）
MCP 配置示例：
{
"mcpServers": {
"visual-understand": {
"command": "uvx",
"args": ["visual-understand-mcp"],
"env": {
"VISION_API_BASE": "https://dashscope.aliyuncs.com/compatible-mode/v1",
"VISION_MODEL": "qwen-vl-max",
"VISION_API_KEY": "sk-xxx"
}
}
}
}安装方式二：本地源码运行
{
"mcpServers": {
"visual-understand": {
"command": "uv",
"args": [
"--directory", "/path/to/visual-understand-mcp",
"run", "visual-understand-mcp"
],
"env": {
"VISION_API_BASE": "https://dashscope.aliyuncs.com/compatible-mode/v1",
"VISION_MODEL": "qwen-vl-max",
"VISION_API_KEY": "sk-xxx"
}
}
}
}配置项
变量 | 说明 | 默认值 |
| 视觉模型 API 地址 | 无（必填） |
| 视觉模型名称 | 无（必填） |
| 视觉模型密钥 | 无（必填） |
| 输出随机性 |
|
| 最大输出长度 |
|
| 请求超时（秒） |
|
| 视觉模型系统提示词 | 见下方 |
前三个必填，其余可选。配置也可写入 ~/.visual-understand-mcp/config.json，env 优先级更高。
默认系统提示词：
你是一个图片分析助手，仅用于理解图片内容。请按以下结构分析图片，根据实际内容调整详细程度：1. 文字内容 — 逐字提取图中所有可见文字，保留原始格式和层级关系。2. 错误信息与代码 — 如果图中包含错误信息或代码片段，必须原样输出，不做任何改写或概括；如果没有则忽略此项。3. 视觉布局与元素 — 描述空间排列、尺寸、颜色及关键元素间的关系。UI 截图请识别组件类型及其状态。4. 数据与指标 — 图表、表格请提取数值、坐标轴、标签、趋势及异常数据点。5. 整体概述 — 概括图片的主题、场景和关键信息，提供完整的上下文理解。不适用的部分跳过。只描述图片中可见的内容，不做推理、猜测或延伸解读。保持精确客观，不推测不可见的内容。
建议写入 CLAUDE.md
进行图片识别任务时，使用 visual-understand MCP 的 understand_image 工具。本地调试
uv run mcp dev src/visual_understand_mcp/server.pyLicense
MIT
Available Tools
1 toolunderstand_imageA
调用视觉模型理解图片，将识别结果以文本返回。
当用户要求识别、分析、OCR、描述、比较图片或截图时，必须调用此工具。 这是唯一可以"看到"图片的工具，不要用 Read/cat 等文本工具打开图片文件。
Args: prompt: 图片理解任务描述，如"提取图中文字"、"分析截图中的报错信息"。 image_path: 单张本地图片路径。 image_url: 单张网络图片 URL 或 data:image base64 格式。 image_paths: 多张本地图片路径列表。 image_urls: 多张网络图片 URL 列表。
ParametersJSON Schema
| Name | Required | Description | Default |
|---|---|---|---|
| prompt | Yes | ||
| image_url | No | ||
| image_path | No | ||
| image_urls | No | ||
| image_paths | No |
Output Schema
ParametersJSON Schema
| Name | Required | Description |
|---|---|---|
| result | Yes |
TDQS
A4.5/5.0
Behavior3/5
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
No annotations are provided, so the description carries full burden. It states that the tool returns text from a visual model, but does not disclose potential limitations (e.g., model accuracy, supported image formats, size limits) or behavioral traits like error handling. The warning against text tools adds some transparency, but overall it's minimal beyond the core function.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Conciseness5/5
Is the description appropriately sized, front-loaded, and free of redundancy?
The description is concise with no wasted words. It front-loads the purpose and usage guidance in a few sentences, then lists parameters in a structured Args block. Every sentence adds value, and the structure is easy to scan.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Completeness4/5
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
Given the tool has an output schema (not shown but indicated), the description does not need to detail return values. It covers the main purpose, usage, and all parameters. However, it lacks details on the visual model's capabilities (e.g., supported image formats, size limits, or failure modes). This is a minor gap but the description is still fairly complete for a tool with no siblings.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Parameters5/5
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
The description has 0% schema description coverage, but the Args section in the description compensates fully. It explains each parameter: prompt with examples ('extract text in image', 'analyze error info in screenshot'), image_path as local path, image_url as URL or base64, and their plural counterparts. This adds meaningful semantics beyond the schema's types and defaults.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Purpose5/5
Does the description clearly state what the tool does and how it differs from similar tools?
The description clearly states that the tool calls a visual model to understand images and returns text. It lists specific use cases (recognition, analysis, OCR, description, comparison) and explicitly distinguishes itself as the only tool that can 'see' images, contrasting with text-based tools like Read/cat.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Usage Guidelines5/5
Does the description explain when to use this tool, when not to, or what alternatives exist?
The description provides explicit guidance on when to use the tool ('when the user asks to recognize, analyze, OCR, describe, or compare images or screenshots') and when not to ('do not use Read/cat and other text tools to open image files'). It asserts that this is the only tool for such tasks, giving clear usage boundaries.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections. Dates show when Glama detected each change.
1 tool update
v0.1.2- First observed
understand_image
TDQS
A4.4/5.0
Disambiguation5/5
Only one tool exists, so there is no possibility of confusion between tools.
Naming Consistency5/5
The single tool 'understand_image' follows a clear verb_noun pattern, which is consistent and intuitive.
Tool Count2/5
With only one tool, the server feels thin for typical use cases; a broader domain like visual understanding might benefit from more specialized tools.
Completeness5/5
The tool's comprehensive description covers recognition, analysis, OCR, description, and comparison of both single and multiple images, leaving no obvious gaps.
Maintenance
ActivityStale
ResponsivenessSyncing
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
Codebase intelligence for agents: 152 structured artifacts across 21 programs, one call.
Turn any LLM multimodal; generate images, voices, videos, 3D models, music, and more.
Production-readiness for your AI coding agents.
Shared memory for coding agents. Stop re-explaining your codebase every session.
Related MCP Servers
- 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.2-
- AlicenseNot gradedqualityCmaintenanceEnables text-only models like Claude Code to recognize images by calling Qwen vision models from DashScope, returning text descriptions for continued reasoning.MIT
- AlicenseAqualityAmaintenanceGives text-only LLM coding agents vision by routing images to a multimodal model and returning detailed textual descriptions. Supports local files, URLs, clipboard, base64, raw bytes, and multiple providers like OpenAI, Anthropic, and Gemini.19912MIT
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/bcdxc/visual-understand-mcp'
If you have feedback or need assistance with the MCP directory API, please join our Discord server