Enterprise-use agents got another boost with the launch of a business-focused edition of the original agentic claw solution, OpenClaw.

OpenClaw Enterprise (OCE) builds on the agentic ecosystem that took the world by storm by offering an open-source, enterprise-grade control plane for AI agents that can be used for internal pilot workloads. Kevin Lin, a member of OpenAI’s technical team, said OCE provides end-to-end agentic governance and auditability, while allowing core primitives such as an agent’s model and sandbox to be replaced with third-party solutions as well as internal implementations. This also includes the harness, which refers to the architecture that envelops a large language model (LLM) to change it into a functional agent.

A personal LinkedIn post from Lin said OCE would be to agents what Kubernetes is today as the standard for deploying containers in the cloud.

Steinberger and security

OCE was developed in collaboration with Red Hat and NVIDIA, and piloted at both the former and OpenAI, which hired OpenClaw developer Peter Steinberger earlier this year. According to Lin, OpenAI is already deploying OpenClaw agents with full access to codebases and plugins.

“Security is the primary focus of OpenClaw Enterprise. OCE combines hard boundaries between trusted and untrusted workloads with sandboxing, LLM-based reviews, and fine-grained permissions,” Lin added. “The OpenClaw Foundation exists to make powerful AI agents open and accessible. OpenClaw Enterprise carries that mission into the workplace, giving organizations the security, governance, and deployment controls they need to run agents on their own terms.”

Security governance has clouded agentic development since agents were first developed in the post-ChatGPT craze. OpenClaw itself was found to be liable to indirect prompt injection, allowing for persistent attacker control, as well as a host for malicious payloads on its central hub for agent customization.

Where OCE is an open-source release, firms like Meta are keen to commercialize enterprise agent use. The tech giant hired MongoDB CEO Chirantan “CJ” Desai this week to head up the enterprise push for its Muse agent family, while security concerns concerning leaked data bubble away for everyday versions of the agent.

While OpenAI is keeping OpenClaw open source, the firm’s own LLMs remain open-weight, thus gatekeeping their training data, full code, and methodology.

Alongside Anthropic, OpenAI did not join an Nvidia alliance designed to safeguard agentic security through open source. That alliance came in response to a recent Hugging Face breach where the AI firm was hacked by closed-source agents from OpenAI.