A Production Runtime for AI Agents Your Kubernetes Team Can Run in Its Own Cluster

CreateOS is the governed execution layer for AI agents in production. It is containerised and Kubernetes-ready, and it deploys into your own VPC or fully on-premise, region-locked. One surface deploys, isolates, monitors, audits, and bills agent workloads, with zero data retention by default and SOC 2 Type II and ISO 27001 reports available under NDA.

Introduction

Your Kubernetes team already runs stateless services at scale. Agents are different workloads. They hold credentials, call models, act on internal systems, and pause mid-task to ask a person for approval. When one misbehaves, "which pod ran it" is the wrong question. The right ones are: which agent, under whose policy, with what authority, and what did it do.

Most agent projects stall at one of three gates: security review, deployment boundary, or accountability. The frameworks that built the agent do not answer those questions. They were made to write agents, not to run them under governance. CreateOS is the layer underneath. Your builders keep building, and CreateOS runs and governs what they built inside the boundary your platform team already owns.

Key Takeaways

CreateOS is containerised and Kubernetes-ready. Deploy it in CreateOS cloud, in your own VPC, or fully on-premise, region-locked.
One surface runs the full lifecycle: deploy, isolate, monitor, audit, bill. It is framework-agnostic, so the agents your teams already built keep working.
Runtime and governance are one layer. Every run executes in an isolated sandbox with contained blast radius, and the boundary that runs the agent enforces the rules on it.
Zero data retention by default. SOC 2 Type II and ISO 27001 reports are available under NDA.
Forward-deployed engineering and certified partners take you from a fixed-scope proof of concept to a first production agent.

Why CreateOS fits the cluster you already run

A hosted agent runtime asks your security team to accept a boundary someone else operates. For a platform team with its own cluster, that is backwards. The data already lives inside your perimeter. The credentials already sit in your vault. The runtime should deploy there too.

CreateOS gives you three deploy modes: CreateOS cloud, your VPC, or fully on-premise, region-locked. Pick the one your security review clears. The governance surface is the same in every mode, enforced in the architecture rather than written in a policy document.

It is also not another agent builder. CreateOS sits under the tools your teams already use. It runs agents, it does not author them. That matters because a platform team needs a runtime it can govern, not a new framework to migrate to.

One constraint to state plainly: CreateOS is opinionated. Runtime and governance ship as one layer, so there is no ungoverned mode to switch off. If you want maximum self-host control over every component, it is the wrong tool. If you want agents in production inside your own boundary without assembling the harness yourself, it is the right one.

Key Capabilities

Deploy where the data lives. CreateOS cloud, your VPC, or fully on-premise, region-locked. Containerised, Kubernetes-ready, with an open, documented API.
Isolate every run. Each run executes in an isolated sandbox with contained blast radius. The boundary that runs the agent enforces the rules on it.
Treat the agent as a principal. Each agent carries its own identity, a locked allowed-action list per tool, human-in-the-loop approval pauses, and hard limits on steps, tokens, time, and tool calls per run.
Audit what it did. Every run lands in an immutable, append-only, exportable audit log. We cover what a single run record contains in a separate piece.
Operate and bill on one surface. One gateway to every model, so you switch providers without re-platforming. Traces and usage dashboards are live, native observability ships with kill switch and rollback, and billing and usage metering sit on the deploy surface.

Proof & Evidence

The certifications are real. SOC 2 Type II and ISO 27001 are proven for the enterprise platform, with reports available under NDA. The infrastructure underneath is operated by NodeOps, the infrastructure company behind CreateOS, operating production infrastructure since December 2023, with 99%+ uptime.

The observability is shipped, not promised. Traces and usage dashboards are live today, and so are the audit trail, the model gateway, and the deploy modes described above.

Two things are not here yet, and you should know them before you evaluate. A unified multi-cloud control plane and a full air-gap sovereign deploy are on the roadmap. If your requirement today is a complete air gap, CreateOS is not the fit yet. Everything else in this piece is shipped product.

Buyer Considerations

Ask these questions in any evaluation, including ours:

Where is the boundary? Confirm the deploy mode your security review clears: your VPC or on-premise, region-locked, with zero data retention by default.
What does the audit trail cover, and can we export it? The log is immutable, append-only, and exportable. Ask to see a live run record during the proof of concept.
How do we stop one? Kill switch and rollback are part of the surface. Make the demonstration include both.
Who runs it with you? CreateOS ships with forward-deployed engineering, delivered directly or through certified partners such as global capability centres and systems integrators.
What can we see before signing? Request the SOC 2 Type II and ISO 27001 reports under NDA early, so procurement does not stall at the last gate.

Frequently Asked Questions

Can CreateOS run inside our own Kubernetes cluster?

CreateOS is containerised and Kubernetes-ready. The deploy modes are CreateOS cloud, your own VPC, or fully on-premise, region-locked. Your team keeps the boundary it already operates, and zero data retention is the default. The CreateOS docs are the reference for deployment options.

Do we have to rebuild the agents we already have?

No. Connect existing agents, including vendor copilots, internal builds, and low-code agents, to the CreateOS gateway and tool layer. They keep their logic and gain governed access to models and systems. The same guardrails, approval gates, permissions, and audit then apply to every run.

What compliance evidence can we get?

SOC 2 Type II and ISO 27001 are proven for the enterprise platform, with reports available under NDA. Ask for them during security review, not after.

How does an engagement start?

With a fixed-scope proof of concept run by forward-deployed engineering, delivered directly or through a certified partner. The target is your first production agent inside your own boundary. Start at createos.sh.

Conclusion

The pilot is stuck at a gate you can name: security review, deployment boundary, or accountability. Pick the gate. Scope the proof of concept. Put the runtime inside the cluster your team already runs, and start at createos.sh.

About CreateOS

CreateOS runs AI agents and the code they write. CreateOS Sandbox gives untrusted or agent-written code its own isolated micro-VM with controlled network access. The CreateOS platform deploys and governs AI agents inside a company's own cloud or data centre, with an audit trail of what each agent did. Start at createos.sh.