Skip to content
agen-x
Security & Governance

Say yes to agents. We own the blast radius.

Your engineers will run agents whether you allow them to or not. agen-x is the one place where the guardrails are on for every agent, and someone else keeps them on. Isolated, fenced, audited, metered.

The agen-x drone raising a glowing neon shield

Is agen-x secure enough for a security engineer to approve?

agen-x enforces governance in the platform, not a PDF: verifiable workspace sandboxing, true tenant isolation, strict network boundaries with allowlists, scoped credentials never written to disk, native SSO and MFA, full audit export, hard budget caps, and ask-a-human interrupts for any runtime. It is not SOC 2 yet. Available now: the architecture doc, an external scan report, the DPA, and a sandboxed org to attack before you sign.

Governance is what the platform is.

Most hosts hope for the best. agen-x enforces isolation, network boundaries, and audit on every agent automatically, and keeps them on across every release. One platform to approve.

Objections Security raises

Security capabilities, and the risk each one mitigates

Verified sandboxing

Mitigates the container-escape surface.

True tenant isolation

Mitigates cross-tenant exposure.

Strict network boundaries

Mitigates exfiltration and stealer payloads.

Scoped, vaulted credentials

Mitigates plaintext-token leaks.

Native SSO and MFA

Mitigates unmanaged access.

Full audit export

Mitigates “we don’t know what it did.”

Hard budget caps

Mitigates uncontrolled spend.

The security kit, available now

Not SOC 2 yet. We start when two customers make it a renewal condition. What a security engineer can verify today: the architecture, an external scan, the DPA, and a sandbox to attack.

Security kit Free with email

The agen-x architecture doc

Isolation, egress, credentials, audit, data flow, and the control mapping — the document that is the agent policy. Forward it to your security engineer.

External scan report

The latest third-party scan of the platform’s public surface. Sent on request.

Ask on the security call →

Vulnerability-disclosure policy

How to report, what we commit to, and how we credit researchers.

Ask on the security call →

DPA + subprocessor list

A standard data-processing agreement and the subprocessors behind it.

Ask on the security call →

Request a sandbox to attack

A sandboxed org for your security engineer to attack for a week before anyone signs. The audit log is the scoreboard.

e.g. an ops agent on a support queue, next month — and who from Security will be attacking it.

Read the architecture. Watch the fence hold.

Get the architecture doc, or ask for a sandboxed org and try to break it first.

Run agents where Security can say yes.

Reply within one business day · priced per org · bring your own keys, never marked up