We could just self-host.

You can. Open-source agents ship the guardrails off by default, for one runtime, with no SSO. Someone has to turn them on, keep them on across every release, and do it again for the next agent. That someone is what you’re buying.

Can we keep our existing model subscriptions?

Yes. Claude Code, Codex, and Copilot run on your own plan. Everything else runs on your own keys. We never touch the model bill.

Isn’t this just another hosted agent tool?

A hosted tool runs one vendor’s agent. agen-x runs every agent your team uses, and the workforce you build next, under one policy and one bill.

What if someone leaves an agent running?

Hard caps pause agents automatically. Paused agents cost disk only.

For teams already running agents

The secure home for your autonomous agents.

Run any agent 24/7 in a workspace Security has already approved. Guardrails on, kept on by us. Your keys, never marked up.

Contact us →

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

The agen-x drone presenting a glowing core of light

What is agen-x?

agen-x is the sanctioned place to run autonomous agents: a managed, governed platform that runs any of 15 agent runtimes 24/7 in isolated workspaces with strict network boundaries, scoped credentials, full audit trails, native SSO, human approval for risky actions, and real-time meters with hard caps. Bring your own keys — never marked up. Security approves the platform once; every agent inherits it, and the same platform grows into a governed agent workforce.

Supported

OpenClaw
Claude Code
Codex
Hermes

Available

GitHub Copilot
Goose
OpenCode
Gemini CLI
Cursor
Qwen
Kimi
DeepSeek
ACP

15 runtimes. One policy for every agent your team runs.

Verified isolationNetwork fencingNative SSOFull audit trailCustom key vaultsReal-time meters

Agents are running unmanaged.

Plaintext tokens on laptops. Forgotten VPS instances. No audit trails.

Open-source agents ship with the guardrails off. Someone has to turn them on, keep them on, and do it again for every new agent. That someone is agen-x.

Security says no. We say yes.

Any agent. One platform. Approved once.

4 runtimes supported, 11 available

One policy for every agent your team runs. Paying teams choose what moves up.

A human in the loop

Any agent can ask a person before a risky step.

Isolation you can verify

Agents cannot escape their workspace or see another tenant.

Strict network boundaries

Agents reach only what you allow. Everything else is denied and logged.

Scoped, rotatable credentials

Keys and tokens are never written to disk.

Your subscription or your keys

Claude Code and Codex on your plan. Everything else on your keys. Never marked up.

A meter you can see and stop

Hard caps pause agents instead of billing them. Paused agents cost disk only.

Native SSO and RBAC

One security review for every agent, not one per tool.

A workforce on the same platform

Templates, channels, and orchestration. Same policy, same bill.

Watch the fence hold.

15 runtimes, each verified on the image it ships.
Live governance: verifiable isolation, network fencing, and audit logs.
Not SOC 2 yet. We start when two customers make it a renewal condition.
Available now: architecture doc, external scan report, DPA, and a sandboxed org to attack.

The 60-second clip: deploy an agent → disallowed request → denied → audit entry → meter → pause.

Objections, answered

Run agents where Security can say yes.

Contact us →

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