Show HN: cMCP, deny an AI agent's tool call and get a signed receipt
cMCP is a new tool that allows users to deny an AI agent's tool call and receive a signed receipt, providing a verifiable audit trail for AI agent interactions.
cMCP is a new tool that allows users to deny an AI agent's tool call and receive a signed receipt, providing a verifiable audit trail for AI agent interactions.
SynthePulse Insight · AI deep readingMembers
Version 1 · 1 source
cMCP is an open-source gateway that moves MCP tool call policy enforcement into a hardware Trusted Execution Environment (TEE), generating verifiable signed TRACE claims for each session to prevent policy tampering or bypass.
cMCP's README points out that software-level MCP governance cannot guarantee that policy packages are not replaced, decisions are not tampered with, or audit logs are authentic. Because hash checks run in the same operating system as the administrator, attackers can exploit supply chain vulnerabilities to flip in-memory decisions or hold signing keys to rebuild the audit chain.
Therefore, the control plane must run in a place that governed processes cannot touch: a hardware TEE.
This report is based on cMCP's GitHub README, which is the project's self-description and has not been independently verified. All functional descriptions are as claimed by the project; actual effectiveness requires testing.
Primary report
Primary source