My servers
Connect your own AI servers for total privacy, zero usage cost and full control.
Why personal servers?
MeigaHub lets you run AI models on your own computers. This means:
- Your data never leaves your network — total privacy
- No cost per token or per use — just the hardware you already have
- Choose from hundreds of open-source models (Llama, Mistral, Qwen...)
What do you need?
To set up a local server you need three things:
- A PC with a graphics card (GPU) with at least 6 GB of VRAM
- MeigaHub Server installed on that PC (available on GitHub)
- Tailscale installed to securely connect your PC with MeigaHub
1 Install Tailscale
Go to the 'My Servers' section in the panel. You'll find your Tailscale connection key. Download Tailscale on your PC and use it to connect to the network.
2 Install MeigaHub Server
Download and run MeigaHub Server on your PC. The server will automatically detect your GPU and available models.
For 7B parameter models you need at least 6 GB of VRAM. For 13B, about 10 GB. For 70B, you'll need 48+ GB.
3 Add the server in MeigaHub
In 'My Servers', click 'Add server' and enter the Tailscale IP address of your PC (something like 100.x.x.x). MeigaHub will detect the available models.
4 Ready! Start chatting
Your server will appear as active in the panel. Now you can go to Chat and start using your assistant with your own hardware.
Server capabilities
Each MeigaHub server can offer up to three capabilities that share the GPU:
- LLM — Chat and text generation (llama.cpp)
- Whisper — Voice-to-text transcription (whisper.cpp)
- Images — Image generation (Stable Diffusion)
If your server is unavailable, MeigaHub detects it automatically. If you have cloud APIs configured, it will use those as fallback. If not, the chat will wait for it to come back.