The security model - policy, grants, mTLS, hash chains

Four layers

1. Policy - deny, allow, ask, enforced on-device
2. Grants - least privilege, time-boxed
3. Transport - mTLS between device, relay and plane
4. Log - hash-chained, anchored daily

Independently verifiable

Each layer can be checked without trusting the vendor: read the policy file, inspect certificates, verify the chain.

Threats covered

Prompt injection, key exfiltration, rogue agents, silent history edits and runaway spend - each maps to a control.

Next

Run the audit verification yourself.

RemoteAgent Online - 100 solutions · Blog · mona.expert · Sign in free