Encrypted and sealed
Every managed password and key is encrypted with AES-256-GCM before it is stored. The keys that encrypt them are themselves sealed, and only Core can unseal them; its seal key is root-owned and read-only.
How Anchor protects the credentials it holds, keeps each part to its own job, controls who can act, and proves what happened.
A privileged access platform is only as strong as its weakest part, so Anchor does not rely on one. Each layer below holds even if another is bypassed.
Every managed password and key is encrypted with AES-256-GCM before it is stored. The keys that encrypt them are themselves sealed, and only Core can unseal them; its seal key is root-owned and read-only.
Web, Compass, Engine and Connect never hold a database address, an encryption key or a signing secret. Engine never carries a session; Connect never changes a credential.
Every connection between Anchor's parts uses TLS 1.2 or newer with both ends verified, even on a single host. Database connections check the server against the deployment's own CA.
Engine and Compass open no port at all. Every service binds one exact address, and only Web may use a privileged port.
Each service runs as its own unprivileged account. Anchor's administrative accounts are locked with no shell, and any attempt to sign in to them is audited and alerted.
Every service runs under systemd protections: a read-only system, no new privileges, private temporary files, kernel and namespace restrictions, and only the capabilities it needs.
Releases are signed and the installer verifies itself before it runs. Packages come from signed apt and dnf repositories, and upgrades roll out signed releases.
Operations on Anchor's own hosts, such as a database switchover, run only when Core has signed them and an administrator approved them within five minutes.
Anchor will not send a credential to a server it cannot identify. Host keys are approved first, and a changed identity stops connections until someone reviews it.
Anchor ships with its own hardening check. It runs every time a host is validated, and an administrator can run it from the console at any time. It fails the validation if a required control is missing, and it verifies, among others:
The result is written as a report you can keep with your evidence.
The audit trail is append-only and cannot be turned off. Each record keeps who acted, under which authority, on what, why, and the result, and is copied to a separate Evidence database that your SIEM reads through read-only views.
Anchor hardens only what it owns. Your firewall, SSH server, operating system patching and time source stay under your standards: Anchor never changes them, and its uninstaller removes the host changes it made.
Ratings are evidence-based; missing information remains visible rather than becoming an assumed pass. Framework mappings help connect technical findings to governance, risk, and compliance reviews.
Insights adds an advisory perspective to findings and curated evidence. Your team retains authority over changes, with policy and review still part of the decision.
Assist’s approach centers on scoped, sanitized information and a path to human help. Approved diagnostics support the investigation and keep the conversation focused on your issue.
Evidence stays linked to actions and the resources they affect. This provides context for operational investigation and assessment work, instead of treating history as a separate task.
Insights runs inside Anchor by default; no record is sent to an external AI service unless an administrator chooses a provider, and then only redacted context travels.
Anchor is designed to support SOC 2 and FedRAMP control requirements. Evidence and control alignment support your compliance work; they do not constitute certification or authorization.