The OpenClaw project has begun work on an enterprise edition, called OCE, to address widespread bans on the agentic AI tool in business settings. OpenClaw, an early "AI harness" that uses large language models to automate tasks like email and calendar management, drew acclaim from developers but also sharp criticism from security analysts. Gartner called the project an "unacceptable cybersecurity risk" for businesses, and China's CERT warned of "extremely weak default security configurations."
Kevin Lin, an OpenAI staffer, said in a post that "actual deployment of persistent agents remains limited" because organizations demand a stronger common security, safety, and governance standard. OCE version 1.0 is designed to meet that need with an enterprise-grade control plane, multi-tenancy, hard security boundaries, and standardized agentic primitives. Governance and auditability span the agent lifecycle, while core components like the harness, model, and sandbox can be swapped with third-party or internal implementations.
Red Hat's Joe Fernandes compared OCE to Red Hat Enterprise Linux and OpenShift, noting that Linux and Kubernetes helped enterprises move to open, industry-standard infrastructure. The OCE GitHub repository explicitly urges developers to think of the project as "Kubernetes for agents." The announcement arrived the same day OpenAI debuted its own "dots" personal agents and the same month Meta released "Muse," a similar automation product for individuals and small businesses, signaling intensifying competition in the agentic AI space.