IS YOUR AI SAFE? THREAT ANALYSIS OF MCP (MODEL CONTEXT PROTOCOL)

Summary from the TLDR Web Dev newsletter

MCP, a standard protocol for connecting tools to large language models, has some vulnerabilities that need protecting against, such as composability chaining and tool poisoning.