Lens Agents governs AI agents with access & policy controls
Lens by Mirantis has come forward with Lens Agents, a governed platform for running AI agents across enterprise systems.
The technology gives organisations a unified, policy-driven way to run, secure and scale AI agents across desktop and cloud environments.
Now in early access, Lens Agents enables developers to connect any AI agent (including desktop tools like Claude, Cursor and Copilot) along with external autonomous agents built on any framework and platform agents that are created directly on Lens Agents.
It enables those entities to connect to enterprise systems with full governance, auditability and policy controls.
The introduction marks an expansion for Lens, beyond the Kubernetes IDE already used by developers, into a broader platform for enterprise AI operations.
Creating centralised control
The company says that the rationale for its launch is the fact that AI agents are rapidly being deployed across engineering, business operations and customer workflows, but mostly outside centralised control.
“AI agents are already operating across every enterprise; the question is whether they are governed,” said Miska Kaipiainen, head of product, Lens. “Cloud platforms govern what runs only on their cloud, while most AI usage today happens on laptops with access across multiple clouds. None of that is properly governed. Lens Agents provides a single platform to govern all of it, any agent, any model, any environment, under the customer’s rules.”
Lens Agents key capabilities include: agent identity and access control so that every agent operates with a distinct identity governed by enterprise policy.
There is also sandboxed execution – Isolated environments prevent unintended actions or lateral movement; server-side credential injection means credentials are never exposed to agents, reducing risk. A full audit trail enables comprehensive logging across all agent interactions and systems; and for policy-based governance, centralised rules exist for access, behaviour and autonomy levels.
Active cost controls means real-time enforcement of usage and spending limits that stops agents when budgets are reached at the organisation, team and individual agent levels.
Beyond a Kubernetes IDE
Lens Agents builds on Lens’ evolution from a Kubernetes IDE into a broader control plane for modern infrastructure and AI workflows.
The announcement follows the launch of the Lens MCP Server last month, which enabled AI assistants to connect to Kubernetes environments. Lens Agents extends that foundation into a full platform for governed AI execution.
Lens Agents helps developers to drive AI adoption while meeting growing regulatory and security requirements.
The platform supports enterprise compliance frameworks, including SOC 2 Type 1 and ISO 27001; audit and oversight requirements are aligned with emerging regulations such as the EU AI Act; and controlled autonomy means teams can define how much independence agents have – from assistive to fully autonomous.
Lens Agents is available now in early access.
