Skip to main content
Glama
rayner-luo

MIMO Image Recognition MCP

by rayner-luo
OverviewSchemaRelated ServersScoreDiscussions
JavaScript
Remote

MIMO Image Recognition MCP

一个调用小米 MIMO 多模态模型进行图片理解的 MCP Server

重要说明

本 MCP 的作用是在不中断 MIMO 2.5 Pro 主模型对话上下文的前提下，通过 MCP 工具调用 MIMO 2.5 模型单独完成图片识别任务。

它不会让 MIMO 2.5 Pro 模型本身具备多模态图片理解能力；图片理解由本 MCP 背后的 MIMO 2.5 模型完成，再把识别结果返回给主对话模型继续推理。

该 MCP 支持：

  • 本地图片路径识别

  • 网络图片 URL 识别

  • Agent 自定义提示词

  • API Key、请求地址、模型名称通过 MCP 启动配置传入

  • 支持通过 npx 运行

  • 支持本地源码运行

Related MCP server: VisionPower

功能说明

本项目会向 MCP 客户端暴露一个工具：

understand_image

用于调用 MIMO 多模态模型理解图片。

支持的输入方式：

  • image_path: 单张本地图片路径

  • image_url: 单张网络图片 URL

  • image_paths: 多张本地图片路径

  • image_urls: 多张网络图片 URL

  • prompt：由 Agent 自己决定的图片理解任务

  • system_prompt：可选系统提示词

  • temperature：输出随机性

  • max_tokens：最大输出长度

建议写入 CLAUDE.md

为了让 Claude 在图片识别、OCR、截图分析等任务中稳定调用本 MCP，建议在项目的 CLAUDE.md 中加入类似说明：

## ⚠️ 模型多模态限制与图像处理规范（核心）

1. **禁止直接处理图片：**
   当前运行的模型（mimo-v2.5-pro）为纯文本模型，**不支持多模态输入**。严禁将任何图片文件（包括截图、本地图片）作为图像输入直接发送给模型，否则会导致系统报错崩溃。

2. **识图任务强制路由至 MCP：**
   当遇到任何需要“看图、识别图片、检查页面视觉布局、UI核对”的任务时（例如使用 `chrome-devtools` 截取了页面），模型**必须且只能**调用 `mimo_image_mcp` 工具。

3. **具体执行链条（以 chrome-devtools 为例）：**
   - 步骤 1：调用工具进行页面访问或截图，将图片保存到本地临时路径（例如 `screenshot.png`）。
   - 步骤 2：**绝对不要**读取图片内容发给模型。
   - 步骤 3：直接调用 `mimo_image_mcp`，将刚才保存的图片本地路径或参数传给它。
   - 步骤 4：接收该 MCP 返回的**纯文本识别报告**，并基于该文本报告回答用户的布局问题。

安装方式一：通过 npx 使用

如果你只是想使用这个 MCP，推荐使用这种方式。

MCP 配置示例：

{
  "mcpServers": {
    "mimo-image-mcp": {
      "command": "npx",
      "args": ["-y", "mimo-image-mcp"],
      "env": {
        "MIMO_API_KEY": "用户自己的 API Key",
        "MIMO_API_BASE": "https://token-plan-cn.xiaomimimo.com/v1",
        "MIMO_MODEL": "mimo-v2.5"
      }
    }
  }
}

配置项说明：

配置项

说明

MIMO_API_KEY

你的 MIMO API Key

MIMO_API_BASE

MIMO API 请求地址，通常为 https://api.xiaomimimo.com/v1或https://token-plan-cn.xiaomimimo.com/v1

MIMO_MODEL

要调用的 MIMO 模型名称，例如 mimo-v2.5

安装方式二：本地源码运行

如果你想修改源码或参与开发，可以使用本地源码方式。

1. 克隆项目

git clone https://github.com/rayner-luo/mimo-image-mcp.git
cd mimo-image-mcp

2. 安装依赖

npm install

3. MCP 配置示例

{
  "mcpServers": {
    "mimo-image-mcp": {
      "command": "node",
      "args": ["<你的项目路径>/src/index.mjs"],
      "env": {
        "MIMO_API_KEY": "你的 MIMO API Key",
        "MIMO_API_BASE": "https://token-plan-cn.xiaomimimo.com/v1",
        "MIMO_MODEL": "mimo-v2.5"
      }
    }
  }
}

请把 <你的项目路径> 改成你自己本地项目的真实路径。

本地调试

可以使用 MCP Inspector 调试：

npm run inspector

如果能看到：

understand_image

说明 MCP Server 启动成功。

如果你想在 MCP Inspector 中实际调用 MIMO 接口，可以在当前终端临时设置：

Windows PowerShell

$env:MIMO_API_KEY="你的 MIMO API Key"
$env:MIMO_API_BASE="https://api.xiaomimimo.com/v1"
$env:MIMO_MODEL="mimo-v2.5"

npm run inspector

License

MIT

Available Tools

1 tool
understand_imageA

Use this tool for ALL image understanding tasks.

This tool calls Xiaomi MIMO multimodal model to inspect and understand images. Whenever the user asks to read, understand, describe, compare, OCR, extract text from, analyze, classify, or answer questions about an image, screenshot, photo, UI capture, chart, poster, document image, receipt, label, or any visual file, you MUST call this tool before giving a final answer.

调用小米 MIMO 多模态模型理解图片。

CRITICAL: This is the ONLY tool allowed to open, read, or see image files (.png, .jpg, .webp). If you have a file path pointing to an image, DO NOT use Read, cat, or any file-reading shell commands. Using standard file-read tools on binary images will cause a system crash. Always use this tool to look at or open an image.

支持单图和多图。Agent 应根据当前任务自己填写 prompt。

Args: prompt: Agent 自己决定的图片理解任务 image_path: 单张本地图片路径 image_url: 单张网络图片 URL 或 data:image base64 image_paths: 多张本地图片路径 image_urls: 多张网络图片 URL system_prompt: 可选系统提示词 temperature: 输出随机性，越低越稳定 max_tokens: 最大输出长度

Returns: MIMO 模型返回的图片理解结果。

ParametersJSON Schema
NameRequiredDescriptionDefault
promptYesAgent 自己决定的图片理解任务
image_urlNo单张网络图片 URL 或 data:image base64
image_pathNo单张本地图片路径
image_urlsNo多张网络图片 URL
max_tokensNo最大输出长度
image_pathsNo多张本地图片路径
temperatureNo输出随机性，越低越稳定
system_promptNo可选系统提示词

TDQS

A4.6/5.0
Behavior4/5

Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?

With no annotations provided, the description fully covers behavioral aspects: it calls the Xiaomi MIMO model, supports single and multi-image, and warns against using file-reading tools. It does not mention rate limits or auth, but the primary safety and behavioral traits are disclosed.

Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.

Conciseness4/5

Is the description appropriately sized, front-loaded, and free of redundancy?

The description is well-structured with clear sections (English, Chinese, warning, args, returns) and is front-loaded with critical information. It is slightly verbose but every sentence serves a purpose, earning its place.

Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.

Completeness5/5

Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?

Given the lack of an output schema, the description includes the return value. It covers purpose, usage, parameter guidance, and warnings comprehensively. For an 8-parameter tool with no annotations, this description is exceptionally complete.

Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.

Parameters4/5

Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?

Schema coverage is 100% with good descriptions. The tool description adds extra context beyond the schema, such as explaining that the agent must fill the prompt and that temperature controls output randomness. This adds meaningful value for correct parameter usage.

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 the tool is for 'ALL image understanding tasks' and lists specific use cases like OCR, classification, description, etc. It explicitly distinguishes itself from file-reading tools, ensuring no ambiguity.

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 explicitly states when to use this tool ('whenever the user asks to...') and provides a critical warning against using other tools on image files to avoid system crash, offering clear guidance on when not to use alternatives.

Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.

Install Server
RayrunBuild, deploy, and operate MCP servers without becoming an infrastructure team.ray.runAd

TDQS

A4.5/5.0
Disambiguation5/5

With only one tool, there is no ambiguity. The tool's purpose is clearly defined and distinct from any other tools.

Naming Consistency5/5

The single tool follows a clear verb_noun pattern (understand_image), consistent and predictable.

Tool Count3/5

One tool is borderline for an image recognition server; while it covers many tasks via a multimodal model, typical MCP servers have 3-15 tools for better granularity.

Completeness4/5

The tool claims to handle all image understanding tasks (OCR, description, analysis, etc.), covering the domain well. Minor gaps like specific image processing are outside scope.

Maintenance

ActivitySlowing
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

Related MCP Servers

  • llm-vision-mcp

    ghbalf
    A
    license
    Not graded
    quality
    C
    maintenance
    An MCP server that enables any LLM to describe images from file paths, URLs, or base64 data by forwarding them to a supported vision provider such as OpenAI, Anthropic, or local Ollama models.
    1,060
    10
    MIT
  • mimo-vision-mcp by jack4862

    mimo-vision-mcpofficial

    jack4862
    A
    license
    A
    quality
    C
    maintenance
    An MCP server that uses Xiaomi MiMo v2.5 multimodal model to provide image recognition capabilities (description, multi-image analysis, OCR, and image info validation) for text-only main models like deepseek-v4-flash, accepting local paths, URLs, file://, and base64 data inputs.
    4
    1
    MIT

Latest Blog Posts

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/rayner-luo/mimo-image-mcp'

If you have feedback or need assistance with the MCP directory API, please join our Discord server