Xybern decides whether every agent action runs, so we secure the layer itself, at every hop, with encryption, isolation, and a signed record you can verify. Your operations are nobody else’s business.
Because Xybern is inline, not alongside, we answer the question every architect asks first: what if Xybern itself fails? Explicitly, by design.
If Xybern is unreachable, actions queue at the boundary. Nothing slips through, and there is no fail-open mode by default.
Workspace keys are scoped at the HSM level. Cross-workspace access returns a hard 403, an architectural impossibility, not a policy call.
The evaluation engine sees the action type, the agent identity, the trust context, and the policy. It never receives the content of the action.
Isolation is the default state at storage, compute, network, and queue, from the moment a workspace is provisioned.
We share our controls, reports, and test results with your security and procurement teams directly. No NDA for the initial architecture review.
Annual third-party audit of security, availability, and confidentiality controls.
Request reportData processing agreements available. EU residency ensures no transatlantic transfer, with right to erasure for non-vault data.
Request DPARegular third-party penetration testing with full remediation tracking.
Request summaryEvery verification is SHA-256 hashed, HMAC-signed, and chain-linked. Tamper-evident by construction, not by policy.
A single agent action is encrypted at every stage, from the call into Xybern, through the evaluation engine, to the vault entry. No plaintext at rest.
Every ingress and egress path uses TLS 1.3 with modern cipher suites. Certificate pinning is available for private deployments.
Data at rest uses envelope encryption with scheduled key rotation. Keys rotate without disrupting existing vault entries.
Selective encryption for sensitive fields within enforcement records, zeroized when the evaluation pipeline completes.
Bring your own keys via your KMS or HSM. Provision, rotate, revoke, and attest, with every key event recorded in the vault.
Tenant isolation is the default, not an add-on. Escalate to dedicated pools, HSM-backed keys, or run the whole layer inside your own infrastructure.
Standard, Enhanced, Dedicated, and Dedicated plus HSM, isolating storage, compute, networking, and queues.
Evaluate policies locally and forward only what you must, keeping enforcement close to your systems.
Full on-premise or air-gapped deployment with no external dependency.
Pin data to the US, EU, or UK, with workspace isolation at every layer.
Authenticate through your identity provider, provision automatically, and restrict how and where people and agents sign in.
Authenticate through Google or your OIDC provider, or email codes.
SAML via most identity providers, including Okta, Entra, OneLogin, and more.
Provision and deprovision users automatically from your directory.
Role-based access with scoped, time-bound permissions for both people and agents.
Restrict workspace access to your private network.
Emergency access requires a named human and a recorded justification, sealed to the vault.
Each authorisation is written once to a hash-chained vault and can be verified on your own, without trusting us.
Every decision produces a signed, offline-verifiable receipt of what was allowed, escalated, or blocked.
Records are chain-linked and append-only. Any later change breaks the chain, which is how tampering is detected.
Verify any record with the public key. No call back to Xybern is required.
Export a scoped, signed pack for any period, agent, or single decision.
Sample audit logWe share architecture documentation, penetration test reports, and SOC 2 Type II directly with security and procurement. No NDA for the initial architecture review.
Xybern