🔒 Privacy-First AI

Your AI. Your Data.
Your Hardware.

McLab supports two ways to keep your trading queries completely private — run your own local LLM, or connect your own GPU server. Your data never has to leave your control.

The Problem with Cloud AI
⚠️ Something to consider

When using cloud AI services, your queries — including stock analysis, portfolio questions, and trading strategies — are processed on third-party servers. Depending on the provider's data policy, this may include query logging or use for model improvement. For privacy-conscious traders, keeping queries local is a practical alternative.

McLab solves this with two complementary features: Private LLM (point the app to your own local model) and Homelab Tunnel (connect your own GPU server). Both are configured in your user profile — no code changes needed.
How It Works
Default (Cloud AI): Your Browser → McLab App → OpenAI / Cloud LLM → Response ↑ Your queries logged here With Private LLM: Your Browser → McLab App → Your Local LLM → Response ↑ Runs on your own machine — nothing leaves With Homelab Tunnel: Your Browser → McLab App → Your GPU Server → Response ↑ Your hardware, your network, your data
Two Ways to Stay Private
🧠
Private LLM
Point the app to your own local model
Run any OpenAI-compatible LLM locally — Ollama, LM Studio, llama.cpp, or vLLM. Register the base URL and API key in your profile. All AI calls route to your machine instead of the cloud.

Works with any model: Qwen, Gemma, Mistral, Llama, and more.
✅ Pros
  • Zero data leaves your machine
  • No API costs
  • Works offline
  • Any compatible model
⚠️ Cons
  • Requires local GPU/CPU
  • Slower than cloud
  • You manage updates
🏠
Homelab Tunnel
Connect your own GPU server
Have a GPU server at home? Register a Cloudflare Tunnel URL in your profile. The app routes your requests directly to your homelab — no port forwarding, no public IP exposure, no DNS changes needed.

Supports Ollama, LM Studio, llama.cpp, vLLM, or any OpenAI-compatible endpoint on your server.
✅ Pros
  • Use powerful home GPU
  • No public IP needed
  • Secure encrypted tunnel
  • Access from anywhere
⚠️ Cons
  • Requires home server
  • Tunnel setup needed
  • Depends on home uptime
How to Set Up
🧠 Private LLM
1
Install Ollama, LM Studio, llama.cpp, or vLLM on your machine and load a model (e.g. qwen3:8b)
2
Note the local URL — typically http://localhost:11434/v1 (Ollama) · http://localhost:1234/v1 (LM Studio) · http://localhost:8080/v1 (llama.cpp) · http://localhost:8000/v1 (vLLM)
3
Go to your profile → Private LLM section → enter the URL, API key, and model name
4
Save — all AI calls now route to your local model
🏠 Homelab Tunnel
1
Install cloudflared on your home server and create a tunnel pointing to your LLM port (e.g. localhost:1234)
2
Cloudflare gives you a tunnel URL like https://your-tunnel.cfargotunnel.com — no DNS changes or public IP needed
3
Go to your profile → Homelab Tunnel section → enter the tunnel URL and model name
4
Save — the app routes your requests through the tunnel to your home GPU
Which Should I Use?
Use Private LLM if you want to run everything locally on the same machine you browse from — laptop or desktop with a decent GPU.

Use Homelab Tunnel if you have a dedicated GPU server at home and want to access it from anywhere — your laptop, phone, or another device — without exposing your home network.

Both can be configured simultaneously. The app uses your private config when available, and falls back to the cloud default otherwise.