deepseek-pp-multimodal-mcp
v0.1.0
Published
Native Messaging MCP host that gives DeepSeek++ OpenAI image analysis and Gemini video analysis tools.
Readme
deepseek-pp-multimodal-mcp
Native Messaging MCP host for DeepSeek++ multimodal tools.
It provides:
vision_status- checks OpenAI and Gemini configuration.analyze_images- analyzes one or more images with OpenAI image input format.analyze_video- analyzes a video with Gemini video input format.
Install the native host for a local DeepSeek++ extension:
npx deepseek-pp-multimodal-mcp install --browser chrome --extension-id <extension-id>For local source development from this repository:
node bin/deepseek-pp-multimodal-mcp.mjs install --browser chrome --extension-id <extension-id>Configure API keys on the DeepSeek++ MCP preset Env field or in the host process environment:
OPENAI_API_KEY=...
GEMINI_API_KEY=...
OPENAI_IMAGE_MODEL=gpt-4.1-mini
GEMINI_VIDEO_MODEL=gemini-2.5-flash
OPENAI_BASE_URL=https://api.openai.com/v1
GEMINI_BASE_URL=https://generativelanguage.googleapis.comWhen used from DeepSeek++, configure these values from the extension Settings page. OPENAI_BASE_URL should point at an OpenAI-compatible /v1 base URL. GEMINI_BASE_URL should point at the Gemini API host; compatible proxies may include the API version in the URL.
The host does not mock model calls. Missing keys, inaccessible media, provider errors, and oversized inline media return explicit MCP errors.
License
MIT
