cMCP is a confidential runtime that intercepts Model Context Protocol tool calls and evaluates them against Cedar policies within a hardware TEE. It provides tamper-evident, signed TRACE Claims to ensure agents cannot bypass enforced tool-use restrictions.
HOW THIS AFFECTS YOU
●
builderYou can secure agentic workflows by enforcing tool-call policies in a hardware-attested environment.
●
policyThis provides a verifiable way to audit and restrict AI agent actions via signed receipts.