An official Model Context Protocol (MCP) Server enabling AI Agents (Claude Desktop, Antigravity, Cursor, Continue.dev) to control local Apple Silicon MLX image generation, query model statuses, run benchmark suites, and monitor Metal GPU telemetry.
- Zero-Touch Dependency Installer: If your Python environment is missing required packages (
mlx,mflux,diffusers,Pillow), the server automatically detects and installs them silently in the background on demand. - Zero-Touch Configuration: No manual terminal setup required for users or agents!
| Tool | Description |
|---|---|
dreambees_rewrite_prompt |
Transform raw user prompt intent into a model-card-optimized prompt via onboard Gemma 4 LLM |
dreambees_generate_image |
Execute local Apple Silicon Metal GPU image generation with FLUX.2 Klein & Sana 2.0 |
dreambees_list_models |
Query available MLX models, download status, and precision |
dreambees_run_benchmark |
Run 10-test benchmark suite or return cached benchmark report |
dreambees_get_metal_diagnostics |
Fetch Apple Silicon Metal VRAM active allocation, peak memory, and cache limits |
npm run build:mcpAdd the following entry to your claude_desktop_config.json:
{
"mcpServers": {
"dreambees-mlx": {
"command": "node",
"args": [
"/Users/bozoegg/Downloads/DreamBeesMLX-main/mcp_server/dist/index.js"
]
}
}
}In your MCP configuration file (mcp_config.json):
{
"mcpServers": {
"dreambees-mlx": {
"command": "node",
"args": [
"/Users/bozoegg/Downloads/DreamBeesMLX-main/mcp_server/dist/index.js"
]
}
}
}Once connected, you can ask your AI Agent:
- "Generate a 512x512 cyberpunk neon bee using FLUX.2 Klein model via DreamBees MLX MCP tool"
- "Run the DreamBees MLX benchmark suite and tell me the average it/s speed"
- "Check the active Apple Silicon Metal VRAM allocation"