MCP server that exposes Hugging Face Spaces as Model Context Protocol (MCP) tools with special support for Claude Desktop (file management, desktop-mode, etc.).
https://github.com/evalstate/mcp-hfspaceStop manually configuring individual AI tools. This MCP server connects Claude Desktop to the entire Hugging Face Spaces ecosystem with a single line of configuration per model.
Most developers spend hours configuring individual AI services, wrestling with API keys, endpoints, and parameter mapping. mcp-hfspace eliminates that entirely. Drop in a space name, and you instantly get:
Image Generation: black-forest-labs/FLUX.1-schnell becomes a tool in seconds. Claude generates images, saves them to your work directory, and can immediately analyze them with its vision capabilities.
Vision Analysis: Point microsoft/OmniParser at a screenshot and get back annotated images plus structured coordinate data. Perfect for UI automation or accessibility testing.
Voice Synthesis: parler-tts/parler_tts turns any text into high-quality speech files, automatically saved and ready for your projects.
Model Conversations: Qwen/Qwen2.5-72B-Instruct gives Claude access to different reasoning approaches - useful for complex problem-solving or getting alternative perspectives.
Claude Desktop's file management is notoriously finicky. This server solves it with "desktop mode":
{
"mcpServers": {
"mcp-hfspace": {
"command": "npx",
"args": [
"-y",
"@llmindset/mcp-hfspace",
"--work-dir=/your/project/assets",
"shuttleai/shuttle-3.1-aesthetic",
"microsoft/OmniParser",
"hf-audio/whisper-large-v3-turbo"
]
}
}
}
That's it. Three powerful AI capabilities added to Claude Desktop in under 30 seconds.
UI Development: Screenshot your app → OmniParser extracts all interactive elements → Claude generates the corresponding HTML/CSS artifact. No more pixel-pushing.
Content Creation: Generate images with FLUX → Use Claude's vision to iterate on them → Convert final descriptions to speech with Parler TTS. Complete multimedia pipeline.
Code Review: Upload error screenshots → Vision model extracts text and context → Claude provides targeted debugging suggestions. Faster than copying stack traces.
Accessibility Testing: Process UI screenshots through vision models → Get structured data about interactive elements → Generate accessibility reports automatically.
Private Spaces: Your custom fine-tuned models work seamlessly with HF_TOKEN authentication.
Specific Endpoints: Need a particular API endpoint? Just append it: Qwen/Qwen2.5-72B-Instruct/model_chat
Multiple Instances: Run different server instances with different work directories and tokens for project isolation.
ZeroGPU Optimization: Automatically handles queuing and quotas - your workflows don't break when HF is busy.
You're already using Claude Desktop for development. This server multiplies its capabilities by 100x without changing your workflow. Instead of "Claude can analyze images," you get "Claude can generate images, analyze them, extract UI elements, convert text to speech, transcribe audio, and chat with specialized models."
Install it once. Access hundreds of AI models. Build workflows that were impossible before.
The 308 stars and active community aren't just numbers - they represent developers who've discovered what happens when AI model access stops being a configuration nightmare and starts being a productivity multiplier.