
This connects Claude to host firmware security checks that run below the OS level. You get five tools: scan_firmware reads BIOS, UEFI, TPM, and Secure Boot state; check_ant_signatures looks for NSA-ANT-class persistence preconditions; attest_firmware generates an HMAC-signed report verifiable at proofof.ai; gate_inference blocks or allows AI operations based on trust state; and list_threat_model shows what attack surface you're defending against. The design philosophy is honest indicator reporting rather than binary clean/compromised verdicts. Reach for this when you need hardware trust attestation before allowing inference on a machine, especially in sovereign AI or high-assurance deployments where firmware-level persistence is part of your threat model.