OpenClaw Enterprise launches a free control plane for AI agents
IA en un minuto newsroom · Editor: Jon Elgezabal

In 30 seconds
The OpenClaw Foundation has unveiled OpenClaw Enterprise, a free, open tool for governing from a single point the AI agents that run continuously inside a company. It offers isolated environments, fine-grained permissions, full auditing and separation between teams, and it runs on the company's own servers. It started at OpenAI, was built with Red Hat and NVIDIA and is still only ready for internal testing.
On September 29, the OpenClaw Foundation, which maintains OpenClaw (an open AI agent platform), unveiled OpenClaw Enterprise, a platform for managing persistent agents (those that work on an ongoing basis) in sensitive environments.
The foundation itself explains the problem it wants to solve: organizations are asking for a common standard for security and governance before they adopt agents, and in most of them the IT department simply bans platforms like OpenClaw.
What it adds is a control plane, the place from which agents are governed, with separation between multiple teams or customers on the same installation, strict security boundaries between what is trusted and what is not, isolated environments, reviews carried out by an AI model and fine-grained permissions. It also makes it possible to audit the entire lifecycle of the agents and to swap the model, the isolated environment and the program that runs the agent for third-party or in-house alternatives.
The project started inside OpenAI, was donated to the foundation and has been developed together with Red Hat and NVIDIA. It is installed on each company's own infrastructure and, according to the foundation, it will always be free for any organization. Red Hat and OpenAI are already testing it internally.
That said, it is not finished yet: the foundation itself considers it suitable only for internal testing, version 1.0 will arrive later this year, and in the coming weeks it will publish a reference architecture with its security measures.
Why it matters · analysis and opinion
If the IT department bans agents for lack of controls, OpenClaw Enterprise tries to take that argument away: isolation, fine-grained permissions, auditing of the whole lifecycle and separation between teams, installed on the company's own infrastructure and at no cost. Having been developed together with Red Hat and NVIDIA, and with Red Hat and OpenAI already testing it internally, it has a backing that few open tools have at birth. But today it is a test bench, not a product: the foundation itself limits it to internal trials and still has to publish the reference architecture with its security measures. Until the stable version arrives, its list of controls works as a template for asking any agent provider what it offers on each point.
Official source: OpenClaw Foundation · Written with the help of AI: how we make the news


