The OpenClaw Foundation announced OpenClaw Enterprise (OCE), an open source platform for managing persistent AI agents in sensitive enterprise environments. The project, originally started at OpenAI and now developed with Red Hat and NVIDIA, is available for self-hosting today ahead of a 1.0 release later this year. Organizations can clone the repository and deploy via Docker Compose for local development or Kubernetes for internal use.
The announcement addresses a direct problem: most IT departments currently ban agentic platforms entirely. The Foundation said the main feedback from organizations is that stronger security, safety, and governance standards are needed before agents can be adopted at scale. OCE is the attempt to solve that without reducing what the agents can do.
What OCE includes
OCE builds on the original OpenClaw platform with an enterprise-grade control plane. It adds multi-tenancy support, hard security boundaries, and standardized agentic primitives. Governance and auditability now span the full agent lifecycle. The harness, model, and sandbox components can be swapped for third-party or internal implementations, keeping the platform vendor neutral.
Security is the primary focus. OCE combines hard boundaries between trusted and untrusted workloads with sandboxing, LLM-based reviews, and fine-grained permissions. A reference architecture showing how these protections work together will be published in the coming weeks.
Early adoption at OpenAI and Red Hat
OpenAI and Red Hat are piloting OCE internally. OpenAI is already running OpenClaw agents with full access to codebases and plugins. RJ Marsan, a member of the technical staff at OpenAI, described the internal agent Androidclaw: "Alert about a broken build? Androidclaw finds the PR and can quickly fix it. Someone gives feedback that the work tab disappeared? Androidclaw responds in like 30 seconds with the link to the SEV."
Marsan added that the agent can trace streaming issues, publish a pull request with video evidence, and merge it. "Kinda game changing. It's smart," he said.
Built in the open
The OpenClaw Foundation prioritized open sourcing OCE early in development to co-build with the community before the 1.0 release. The project will remain free for any organization to use and runs on an organization's own infrastructure. The Foundation said developers, operators, security teams, and organizations are invited to contribute to the open foundation for enterprise agents.
For IT leaders evaluating agent deployment, the shift from blanket bans to controlled pilots is the core signal here. Teams responsible for governance and security standards may want to track OCE's reference architecture release. Those building internal AI capabilities can explore AI Agent Courses to understand agentic workflows, while executives shaping policy can review AI IT Strategy Training for guidance on deployment controls.
Why this matters for executives and IT leaders
The default stance of banning agentic platforms is becoming harder to sustain as competitors pilot them internally. OCE gives organizations a concrete option to test persistent agents under security controls they manage themselves, on their own infrastructure, without vendor lock-in. The open source model means security teams can inspect the code before deployment, and the multi-tenancy design supports gradual rollout across departments rather than an all-or-nothing decision.
Your membership also unlocks:
