OpenClaw Enterprise (OCE) is an open-source, vendor-neutral platform for running persistent AI agents inside companies with strict security needs. It was announced on 29 September 2026 in a blog post by Kevin Lin, and on 23 October OpenClaw shared a podcast interview in which Kevin explains why he built it and how to get involved. The project’s own README sums it up in one line: “Think of it as Kubernetes for agents.” Here’s what that means, what’s inside, who it’s for, and how to try it — in plain English, with the exact setup steps from the repo.
Key takeaways
- OCE = an open-source control plane for persistent agents, self-hosted on your own infrastructure.
- Built for multi-tenancy, governance and auditability.
- Harness, model and sandbox are replaceable.
- Pre-1.0: “intended for internal pilot workloads”.
- Want agents running safely in your business? Claim a free strategy session.
What is OpenClaw Enterprise?
OpenClaw showed what a good model plus a harness built for agent work can do. But according to the announcement, real deployment of persistent agents inside organisations has stayed limited. The feedback OpenClaw kept hearing was that a stronger shared standard for security, safety and governance was needed first — and that, as a result, many IT teams simply ban agent platforms like OpenClaw.
OpenClaw Enterprise is the answer to that. The blog frames the question as how to safely deploy powerful agents in enterprise environments without weakening what they can do. OCE adds an enterprise-grade control plane on top of the OpenClaw foundation, with:
- Multi-tenancy — separate teams, each with their own agents.
- Hard security boundaries between trusted and untrusted workloads.
- Standardised agent primitives — harness, model and sandbox.
- Governance and auditability across the whole agent lifecycle.
The control plane itself is called the OpenClaw Control Plane (OCC). It’s what you use to deploy and manage Agents.
Why “Kubernetes for agents”?
Kubernetes is the standard way companies run lots of software containers across their own servers, with one central place to deploy, manage and control them. OCE aims to do the same job for agents. The Enterprise page shows it as one control plane on your infrastructure, with Team A and Team B each running persistent agents, and each agent built from three replaceable parts: a harness, a model and a sandbox.
That last part matters. The blog says core primitives “can be swapped out with third party solutions as well as internal implementations.” So you’re not locked into one model provider or one sandbox. That’s what “vendor neutral” means here.
Want persistent agents running safely in your business? Inside the AI Profit Boardroom you get the Agent OS — Hermes, OpenClaw, OpenCode, Claude and Codex in one place — plus step-by-step Hermes training and four live coaching calls a week.
Want a plan for your business first? Claim your free strategy session.
Join the AI Profit Boardroom →Claim Your Free SEO Strategy Session →
Security in OpenClaw Enterprise
The blog calls security “the primary focus” of OCE. It combines four things:
- Hard boundaries between trusted and untrusted workloads
- Sandboxing
- LLM-based reviews
- Fine-grained permissions
The Enterprise page puts it as “boundaries, not just guardrails”: multi-tenancy and hard security boundaries separate workloads, while sandboxing and fine-grained permissions help define where agents can act. The team says a reference architecture showing how these protections work together is coming in the following weeks.
In the code, two packages handle much of this, according to the README’s code layout:
| Path | Responsibility |
|---|---|
| packages/iam/ | Identities, roles, and resource authorization |
| packages/audit/ | Audit events and sensitive-value sanitization |
| apps/controller/ | HTTP API, browser console, worker, and Drivers |
| packages/occ/ | Resource lifecycle, persistence, and work queue |
| cmd/occ/ | Go entry point for the OCC domain CLI |
Who is behind OpenClaw Enterprise?
The blog says the project “originally started at OpenAI and then was donated to the OpenClaw Foundation, where it is now an independent project and has been further developed in collaboration with Red Hat and NVIDIA.” It also says OCE is being piloted internally at companies like Red Hat and OpenAI.
In the post, an OpenAI engineer describes their internal agent triaging chat channels, tracing problems and opening fixes with full context from git, GitHub and logs. That’s the kind of persistent, deeply connected agent OCE is built to govern.
OCE runs on your own infrastructure, and the blog says it “will always be free for any organization to use.” It’s MIT licensed.
Is OpenClaw Enterprise ready for production?
Not yet. The Enterprise page says: “Pre-1.0 software, intended for internal pilot workloads.”
The team chose to open-source it early to build in the open with the community, with a 1.0 release planned for later this year. So the right move today is a pilot: one team, a few agents, non-critical work, and close review. OpenClaw also runs bi-weekly office hours for developers, operators, security teams and organisations exploring the platform.
Want this set up for you instead of doing it yourself? Claim your free strategy session →
How to get started with OpenClaw Enterprise
You can self-host OCE by cloning the openclaw-enterprise repo. The README offers two paths: Local Setup to run OCC on your machine, or Kubernetes Setup to install it on a cluster you already operate.
You’ll need these tools installed first:
- Docker Engine or Podman
- k3d, kubectl and Helm
- Bash and Python 3
- Go (the version in
go.mod) - Node.js 24.22.1 or newer
- pnpm (the version in
package.json)
For local setup, run this from the repository root:
pnpm cli:build
export OCC_DEVELOPMENT_COMPUTE_DRIVER=kubernetes
export OCC_DEVELOPMENT_CONTROL_PLANE=kubernetes
export OCC_DEVELOPMENT_SANDBOX_DRIVER=none
./bin/occ dev upThose exports select the Kubernetes-only profile for deploying Agents. The README warns that if you skip them, occ dev up starts a Compose control-plane preview that cannot deploy Agents.
Next, deploy your first Agent and send it a model request. For that step you need an OpenAI API key with access to the default model (or the override you choose). If you installed on an existing cluster, follow the repo’s guide to deploy and verify an Agent there instead.
Once that works, use Agent Presets to reuse settings: pick a preset, fill in the variables and review the copied draft in the browser console.
OpenClaw Enterprise vs personal OpenClaw
| OpenClaw | OpenClaw Enterprise | |
|---|---|---|
| Built for | Running a powerful agent yourself | Running many persistent agents across an organisation |
| Main idea | A good model plus a harness built for agent work | A control plane (OCC) that deploys and manages Agents |
| Teams | Single user or small setup | Multi-tenancy — separate teams and workloads |
| Controls | Your own setup | Governance, auditability, sandboxing, fine-grained permissions |
| Where it runs | Wherever you install it | Your own infrastructure, deployed via Kubernetes |
| Status | Established | Pre-1.0, for internal pilots |
If you’re a solo founder or a small business, personal OpenClaw (or Hermes) is still the place to start. OCE is for organisations whose IT or security team needs central control before agents are allowed in. If you’re weighing up agents, read my comparison Hermes Agent vs OpenClaw, and for a fun OpenClaw build, OpenClaw Discord Voice.
My verdict on OpenClaw Enterprise
OCE goes after the real blocker for agents at work: not what they can do, but whether a company can trust and control them. Open source, self-hosted, vendor-neutral and backed by contributors including Red Hat and NVIDIA, it has a serious start. It’s pre-1.0, so treat it as a pilot. If your organisation wants persistent agents but IT keeps saying no, OpenClaw Enterprise is the project to watch and test now.
About Julian Goldie
I’m Julian Goldie — founder of Goldie Agency and the AI Profit Boardroom. I test new AI agents almost every day for 432K+ YouTube subscribers and run Hermes, OpenClaw, OpenCode, Claude and Codex side by side in my own Agent OS. Claim your free strategy session or join the AI Profit Boardroom.
Want another angle? Read the AI Profit Boardroom guide to OpenClaw Enterprise.
FAQ: openclaw enterprise
What is OpenClaw Enterprise?
OpenClaw Enterprise (OCE) is an open-source, vendor-neutral platform for managing persistent agents in sensitive environments. It includes the OpenClaw Control Plane (OCC) for deploying and managing Agents.
Is OpenClaw Enterprise free?
The announcement says OCE is built to run on your own infrastructure and will always be free for any organization to use. It is MIT licensed.
Is OpenClaw Enterprise production-ready?
Not yet. It is pre-1.0 software intended for internal pilot workloads, with a 1.0 release planned for later this year.
What do I need to run OpenClaw Enterprise locally?
Docker Engine or Podman, k3d, kubectl, Helm, Bash, Python 3, Go, Node.js 24.22.1 or newer and pnpm. Deploying your first Agent also needs an OpenAI API key.
Who is contributing to OpenClaw Enterprise?
It started at OpenAI, was donated to the OpenClaw Foundation, and has been further developed in collaboration with Red Hat and NVIDIA.
Ready to put Hermes to work in your business? Join the AI Profit Boardroom for the Agent OS, setup training and daily updates.
Or claim your free strategy session and we’ll map out your setup.
Starting from zero? Get the free AI course and community.
Join the AI Profit Boardroom →Claim Your Free SEO Strategy Session →Get the FREE AI Course + Community →
Related reading
Last updated October 2026. This page is a living guide to openclaw enterprise — the facts here move fast and it is updated as they do.
Done-for-you AI SEO
Rank #1 with AI SEO and get more leads, traffic and sales
Done-for-you SEO packages: backlinks, AI-powered articles and an account manager every month. Save 30% with code BLACKFRIDAY (ends Dec 1). No setup fees.