reasonix-vision-mcp

# Tool List Vision Bridge MCP Server — Enables text models to "see" images by calling visual models through MCP. Supports any OpenAI-compatible visual API (SiliconFlow / GPT-4o / Qwen-VL, etc.), including image compression functionality. > ⚠️ **Note:** This project is based on the **SiliconFlow** platform and tested in **Reasonix**. Other platforms/clients have not been thoroughly tested, and compatibility issues may exist. ## Tools | Tool | Description | |------|------| | `describe_image` | Describes image content. Supports local image paths or public URLs | | `compress_image` | Compresses local image files | | `v` | Version information (only for viewing, not for calling) | ### describe_image Parameters | Parameter | Required | Description | |------|------|------| | `image` | ✅ | Image path: local absolute path or public URL | | `prompt` | ❌ | Custom description guide, e.g., "Analyze this UI interface" | | `detail` | ❌ | Visual precision, default is auto. Passes through to API as-is | ### compress_image Parameters | Parameter | Required | Default Value | Description | |------|------|--------|------| | `image` | ✅ | — | Absolute image path | | `maxWidth` | ❌ | `800` | Maximum width (pixels). Only takes effect if maxSize is not specified | | `maxSize` | ❌ | — | Target size, e.g., `"1M"`, `"500KB"`, `"200K"` | When `maxSize` is specified, the tool will compress the image to within the target size using a binary search approach, without exceeding it. ## Installation ### Method 1: npx (Recommended) ```json { "mcpServers": { "reasonix-vision-mcp": { "command": "npx", "args": ["-y", "reasonix-vision-mcp"], "env": { "VISION_API_KEY": "sk-your-key", "VISION_BASE_URL": "https://api.siliconflow.cn/v1", "VISION_MODEL": "Qwen/Qwen3.5-35B-A3B", "VISION_DETAIL": "auto" } } } } ``` ### Method 2: Direct Execution ```bash node dist/src/index.js ``` ## Environment Variables | Variable | Required | Default Value | Description | |------|------|--------|------| | `VISION_API_KEY` | ✅ | — | API Key for the visual model | | `VISION_BASE_URL` | ❌ | `https://api.openai.com/v1` | OpenAI-compatible API address | | `VISION_MODEL` | ❌ | `Qwen/Qwen3.5-35B-A3B` | Visual model name | | `VISION_DETAIL` | ❌ | `auto` | Visual precision, default is auto. Passes through as-is, without validation | | `VISION_OUTPUT_DIR` | ❌ | `./vision-output` | Output directory for compress_image | ## License MIT