Skip to content
Future Site

News

Announcements about open-source releases, new features and company updates.

Thumbnail: 'Is your MCP server safe?' next to a diagram of a prompt injection attack
Tutorial

MCP security: prompt injection, OAuth and least privilege

How AI agents get tricked through MCP servers, and how to stop it: the lethal trifecta, prompt injection via tool output, tool poisoning, OAuth for remote servers and a safe-server checklist.

AI agentsMCPOAuthSecurity
6 min read