Skip to main content
Glama
RunhuaHuang

VisionPower

by RunhuaHuang
OverviewSchemaRelated ServersScoreDiscussions
JavaScript
Hybrid

VisionPower

VisionPower is a portable image-understanding MCP server for Codex, Claude Desktop, Cursor, Cline, and other MCP-compatible agents.

It exposes one tool:

  • describe_image: analyze a local image path, image URL, or base64 image with an OpenAI-compatible vision model.

The default provider is Qwen VL through DashScope's OpenAI-compatible endpoint, but you can point it at any compatible vision API, such as OpenAI gpt-4o.

Requirements

  • Node.js 18 or newer

  • A vision-capable OpenAI-compatible API key

Related MCP server: vision-mcp

Configuration

Environment variables:

Name

Required

Default

Description

RUN_VISION_API_KEY

Yes

API key for the configured vision provider.

RUN_VISION_MODEL

No

qwen3-vl-flash

Vision model name.

RUN_VISION_BASE_URL

No

DashScope compatible endpoint

Base URL without /chat/completions.

RUN_VISION_ALLOWED_DIRS

No

unrestricted

Comma-separated directories local image_path may read from.

RUN_VISION_MAX_IMAGE_BYTES

No

20971520

Max local image size in bytes.

RUN_VISION_TIMEOUT_MS

No

60000

Upstream API timeout.

RUN_VISION_MAX_TOKENS

No

2048

Max response tokens.

Codex

Add this to ~/.codex/config.toml:

[mcp_servers."visionpower"]
type = "stdio"
command = "npx"
args = ["-y", "github:RunhuaHuang/VisionPower#v0.1.1"]

[mcp_servers."visionpower".env]
RUN_VISION_API_KEY = "your-api-key"
RUN_VISION_MODEL = "qwen3-vl-flash"
RUN_VISION_BASE_URL = "https://dashscope.aliyuncs.com/compatible-mode/v1"
RUN_VISION_ALLOWED_DIRS = "/Users/you/Desktop,/Users/you/Downloads"

Claude Desktop / Cursor / Cline

Add this to the app's MCP configuration:

{
  "mcpServers": {
    "visionpower": {
      "command": "npx",
      "args": ["-y", "github:RunhuaHuang/VisionPower#v0.1.1"],
      "env": {
        "RUN_VISION_API_KEY": "your-api-key",
        "RUN_VISION_MODEL": "qwen3-vl-flash",
        "RUN_VISION_BASE_URL": "https://dashscope.aliyuncs.com/compatible-mode/v1",
        "RUN_VISION_ALLOWED_DIRS": "/Users/you/Desktop,/Users/you/Downloads"
      }
    }
  }
}

OpenAI Example

{
  "RUN_VISION_API_KEY": "your-openai-api-key",
  "RUN_VISION_MODEL": "gpt-4o",
  "RUN_VISION_BASE_URL": "https://api.openai.com/v1"
}

Tool Input

{
  "image_path": "/absolute/path/to/image.png",
  "prompt": "Read the text in this screenshot and summarize it."
}

Alternative inputs:

{
  "image_url": "https://example.com/image.png",
  "prompt": "What is in this image?"
}
{
  "image_base64": "...",
  "image_mime_type": "image/png",
  "prompt": "Extract all visible text."
}

Local Development

npm install
npm run smoke

Start the server directly:

npm start

After npm publication, you can replace the GitHub install target with:

["-y", "visionpower"]

Security Notes

image_path lets an agent ask this MCP server to read local raster image files. Set RUN_VISION_ALLOWED_DIRS when giving this MCP to agents you do not fully trust.

The server validates absolute paths, file type, file signature, and file size before sending image data to the configured vision provider.

This server cannot be installed

A
license - permissive license
-
quality - not tested
B
maintenance

How are these scores calculated?

Maintenance

–Maintainers
–Response time
–Release cycle
1Releases (12mo)
Commit activity

Resources

Unclaimed servers have limited discoverability.

Looking for Admin?

If you are the server author, to access and configure the admin panel.

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/RunhuaHuang/VisionPower'

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