Is your LLM server open to the whole internet?
Attackers scan for open Ollama, vLLM and MCP servers, then resell access to them. Run our free check against your own server in one minute, or let us watch it every day and alert you the moment it opens up.
Why this matters now
- In the Cloud Security Alliance's May 2026 note, 31% of 5,200+ Ollama servers queried answered prompts with no authentication, and GreyNoise logged 91,403 attack sessions against exposed LLM endpoints. Source
- In "Operation Bizarre Bazaar", attackers scanned for open Ollama, OpenAI-compatible and MCP endpoints and resold access, about 972 attacks a day. Source
Free check: run it yourself
One Python file, standard library only, read-only requests. Run it from outside your network (a laptop on a phone hotspot works) against a server you own:
curl -O https://inferprice.pages.dev/llm-exposure-check.py
python3 llm-exposure-check.py http://YOUR-SERVER:11434It checks five things: an Ollama API with no auth, an OpenAI-compatible API (vLLM, llama.cpp, SGLang) with no key, CORS open to any website, public /metrics, and an MCP server that accepts initialize with no token. Read the code.
Daily monitoring with alerts: $9 a month
We run the same checks against your endpoints every day from outside your network and email you the moment one fails, for example after a deploy that drops the auth proxy or opens CORS. Up to 5 endpoints. You verify each one is yours before we check it.
Early access: monitoring starts within 30 days of your pre-order. If it doesn't, or you're not happy in the first 30 days, you get a full refund. We only check endpoints you prove you own.