
OpenShell is an open-source runtime that lets developers run autonomous AI agents under enforced, auditable policies.
It runs each AI agent in an isolated sandbox with no direct network access, limits the files the agent can reach, and monitors and filters its system calls in the kernel, while a supervisor outside the sandbox checks every network request against policy and supplies credentials only where policy allows. Sandboxes run through Docker, Podman, or Kubernetes, and deployments can be local, on-premises, hybrid, or in cloud infrastructure.
It is documented as a runtime beneath agent frameworks and harnesses rather than a framework itself, adding gateway-brokered credentials, policy-enforced egress, inference routing, and audit logging that a container or Kubernetes pod alone does not provide. Its named users are developers building autonomous agents, platform teams enabling those developers, and security or IT teams that need to govern agent execution.


AgentShield is an AI agent security control layer that inspects and governs agent traffic in the request and action path.


Rein is an agentic security platform whose data model remains fully in-org.


Platform that discovers, monitors, and governs enterprise AI agents at runtime.
AgentShield is an AI agent security control layer that inspects and governs agent traffic in the request and action path.

Rein is an agentic security platform whose data model remains fully in-org.

Platform that discovers, monitors, and governs enterprise AI agents at runtime.




Open-source Python framework for building sandboxed reinforcement learning environments that run and score AI agents.



Kyvvu's Agent Security Kernel secures AI agents by evaluating each action against task history inside the agent process.


Gatewatch is Rig Security’s endpoint protection product for identifying and controlling AI agent activity.


Dataiku Agent Management inventories, monitors, and governs AI agents across enterprise platforms.



Platform securing machine learning models and AI systems against attacks.


Security platform for protecting generative AI and LLM applications in enterprises.



bVisor is a lightweight embedded bash sandbox runtime for AI agents.

Open-source Python framework for building sandboxed reinforcement learning environments that run and score AI agents.

Anjuna Overwatch is a trusted control plane for autonomous AI workloads.

Kyvvu's Agent Security Kernel secures AI agents by evaluating each action against task history inside the agent process.

Gatewatch is Rig Security’s endpoint protection product for identifying and controlling AI agent activity.

Dataiku Agent Management inventories, monitors, and governs AI agents across enterprise platforms.

Arx is Fabraix's run-time defence layer that blocks prompt injection and memory poisoning in AI agents.

Platform securing machine learning models and AI systems against attacks.

Security platform for protecting generative AI and LLM applications in enterprises.

TrustGate connects agents to models and tools securely as part of NeuralTrust.

Lakera Guard is a runtime security product for large language model applications.

Zenity is an AI agent security and governance platform that governs how autonomous agents behave.
