Verifying Humans and Agents at the Point of Action
Credentials cannot distinguish a human from an AI agent holding the same token: a session token, API key or delegated grant proves authorization, not who is operating it at the moment of use. MagenTrust continuous identity assurance evaluates live signals before each consequential action and returns a deterministic PASS, CHALLENGE or INTERRUPT. MagenMCP applies that decision to agent tool calls in the Model Context Protocol path, before execution. Inside a Human Exclusion Zone the model inverts, so any human signal returns INTERRUPT and no human signal returns PASS, giving machine-only execution you can prove. Humans and machines run under one policy, inside your perimeter, with every decision logged as audit evidence.