Ollama Scout

Probe internet-exposed Ollama endpoints: list installed models, read VRAM/GPU usage of loaded models, and send a live test prompt to confirm the LLM responds.

The "Vulnerability"

Not really a true vulnerability in the sense it could be that damaging, but it allows anyone to use your Ollama server or VPS to run AI on, if it is exposed to the internet, like a lot of VPS's are.

Run

pip install -r requirements.txt
python app.py
# then open http://localhost:8787

Change the port with PORT=9000 python app.py.

Finding targets with Shodan

Shodan is a search engine for internet-connected devices.

1. Go to shodan.io and create a free account.

2. In the search bar, enter:

   product:Ollama
   

3. Browse the results. Each result shows an IP address. Ollama exposes its API on port 11434 by default.

4. Copy an IP (e.g. 1.2.3.4 or 1.2.3.4:11434) and paste it into the Ollama Scout input field, then hit Scan host.

Other useful Shodan queries:

Query	What it finds
product:Ollama	All detected Ollama instances
product:Ollama country:US	Filter by country
product:Ollama org:"Amazon"	Filter by hosting provider

AI Usage

• Grok 4.5 used in public/index.html
• DeepSeek v4 Flash used in .gitignore