§ 01Raptoric Research

Original research on the security of AI

Data-led reports on the systems that are being built faster than they are being secured: AI agents, the Model Context Protocol, and the tools they run on. Every report is grounded in cited evidence, honest about what the data does and does not prove, and free to read and cite.

Raptoric ResearchEdition 1.0 · June 2026 · 22 min read

The MCP and AI-Agent Attack Surface

Tool-using AI shipped its convenience first and its security model second. This is the evidence on where it breaks, what the data does and does not prove, and how to defend agents that run on attacker-controlled input.

Read the report