SiliconFlow MCP Server

[English | 한국어]

An MCP (Model Context Protocol) server for SiliconFlow's generative services. This allows AI models (like Claude) to generate high-quality images, videos, and speech directly using various models available on SiliconFlow.

• Service: SiliconFlow
• API Documentation: SiliconFlow API Reference

Features

• generate_image tool: Generate images from text prompts.
  • Supports multiple models (FLUX.1-schnell, FLUX.1-dev, FLUX.2-pro, etc.)
  • aspect_ratio support: Choose from 1:1, 16:9, 9:16, etc.
  • Supports negative_prompt for supported models.
  • Customizable seeds for reproducible generations.
• generate_video tool: Generate videos via text prompts (auto-polls until completion).
  • Supports Wan-AI models and customizable aspect ratios.
• generate_speech tool: Generate speech (TTS) from text.
  • Supports fish-speech, IndexTTS, and CosyVoice models.
  • Customizable voices, response formats (mp3, wav, etc.), and speed.
• submit_video_generation & get_video_status: Low-level tools for manual async video management.
• list_models tool: Dynamically fetch available image, video, and audio models.
• get_user_info tool: Check your SiliconFlow account details, including balance (Total, Paid, Free) and profile info.
• Local Saving: Automatically save .png, .jpg, .mp4, or .mp3 files to your specified directory.

Setup

1. Prerequisites

• uv installed (recommended) or Python 3.10+.
• A SiliconFlow API Key. Get one at SiliconFlow Dashboard.

2. Configuration

The server requires an API key to function. You can provide it via environment variables or a .env file.

SILICONFLOW_API_KEY=your_api_key_here
# Optional: Path to save generated images/videos locally
SILICONFLOW_IMAGE_DIR=/path/to/save/assets
# Optional: Path to save generated audio files (defaults to IMAGE_DIR)
SILICONFLOW_AUDIO_DIR=/path/to/save/audio

Usage

Using with uvx (Recommended)

You don't need to install anything locally. Just run it directly using uvx:

uvx siliconflow-mcp

Installation via PyPI

You can also install it as a global tool:

uv tool install siliconflow-mcp
# or using pip
pip install siliconflow-mcp

Configuration for MCP Clients

Claude Desktop

Add the following to your Claude Desktop configuration file (%APPDATA%\Claude\claude_desktop_config.json on Windows or ~/Library/Application Support/Claude/claude_desktop_config.json on macOS):

{
  "mcpServers": {
    "siliconflow": {
      "command": "uvx",
      "args": ["siliconflow-mcp"],
      "env": {
        "SILICONFLOW_API_KEY": "your_api_key_here",
        "SILICONFLOW_IMAGE_DIR": "/path/to/save/assets",
        "SILICONFLOW_AUDIO_DIR": "/path/to/save/audio"
      }
    }
  }
}

Claude Code

Run the following command:

claude mcp add siliconflow \
-e SILICONFLOW_API_KEY="your_api_key_here" \
-e SILICONFLOW_IMAGE_DIR="/path/to/save/assets" \
-e SILICONFLOW_AUDIO_DIR="/path/to/save/audio" \
-- uvx siliconflow-mcp

Gemini CLI

Run the following command:

gemini mcp add siliconflow \
-e SILICONFLOW_API_KEY="your_api_key_here" \
-e SILICONFLOW_IMAGE_DIR="/path/to/save/assets" \
-e SILICONFLOW_AUDIO_DIR="/path/to/save/audio" \
uvx siliconflow-mcp

Manual: Add the configuration to your settings.json (usually located in .gemini/settings.json):

{
  "mcpServers": {
    "siliconflow": {
      "command": "uvx",
      "args": ["siliconflow-mcp"],
      "env": {
        "SILICONFLOW_API_KEY": "your_api_key_here",
        "SILICONFLOW_IMAGE_DIR": "/path/to/save/assets",
        "SILICONFLOW_AUDIO_DIR": "/path/to/save/audio"
      }
    }
  }
}

Installation for Developers

If you want to contribute or run from source:

# Install dependencies
uv sync

# Run the server locally
uv run siliconflow_mcp

Supported Models

Image Models

• black-forest-labs/FLUX.1-schnell (Fast and efficient)
• black-forest-labs/FLUX.1-dev (High fidelity)
• black-forest-labs/FLUX.1-pro (Top-tier quality)
• stabilityai/stable-diffusion-3-5-large
• stabilityai/stable-diffusion-3-5-large-turbo
• stabilityai/stable-diffusion-xl-base-1.0
• ByteDance/SDXL-Lightning
• Kwai-Kolors/Kolors
• Qwen/Qwen-Image-Edit-2509 (Image editing)

Video Models

• Wan-AI/Wan2.2-T2V-A14B (Text-to-Video)
• Wan-AI/Wan2.1-T2V-14B
• Wan-AI/Wan2.1-I2V-14B-720P (Image-to-Video)
• Wan-AI/Wan2.1-T2V-1.3B

Audio (TTS) Models

• fishaudio/fish-speech-1.5
• IndexTeam/IndexTTS-2
• FunAudioLLM/CosyVoice2-0.5B

You can use the list_models tool to see the full list of available models from SiliconFlow.

License

MIT