Glossary
Plain-language definitions for the terms being coined in real time — prompt injection, tool poisoning, rug pulls, non-human identity, toxic flow, and the protocol plumbing underneath them.
Open the glossary →The vocabulary and the exposure map of agentic AI security.
A plain-language field guide to agentic AI and MCP security — plus a passive exposure map of internet-facing MCP servers and LLM gateways.
- Terms defined
- 46
- Exposed instances tracked
- 0
- Candidates identified
- 71
- Updated
- 2026-09-08
Exposure Map
A passive, fingerprint-only census of internet-facing MCP servers and LLM gateways. No authentication, no tool invocation, no data extraction — and an opt-out that is honored.
View the map →