Lens Launches MCP Server to Connect AI Coding Assistants with Kubernetes

Lens by Mirantis announces a built in Model Context Protocol server in Lens Desktop, expanding how teams connect AI coding assistants to Kubernetes in a secure and straightforward way. The addition makes Lens a practical bridge between popular AI tools and real infrastructure, so software delivery and operations tasks can be handled within existing workflows rather than through custom scripts or one off plug ins.

Now developers and platform engineers using AI coding assistants including Claude Code, ChatGPT, Cursor, GitHub Copilot can more easily discover, connect to, and manage Kubernetes clusters with a built in MCP server added to Lens, the worlds most widely adopted Kubernetes IDE with more than 1 million users worldwide.

A simpler path from AI assistants to Kubernetes

AI coding assistants are becoming part of daily development, from writing code to fixing issues as they appear. Until now, bringing Kubernetes context into those tools has often required extra setup, ad hoc scripts, and managing kubeconfig inside the AI clients. With the built in MCP server in Lens Desktop, AI tools can request cluster access and context from Lens, which already manages connectivity for the user. This reduces setup friction and centralizes access in one place that developers and platform engineers know and trust.

How the built in MCP server works in Lens Desktop

The Lens MCP server exposes a standard interface that MCP compatible assistants and agents can use to discover clusters a user has configured in Lens. Instead of shipping custom adapters for each AI tool, Lens provides a consistent connection point. When an AI assistant asks to list clusters, retrieve contexts, or run kubectl style queries, Lens supplies that capability using the same authentication and permissions that the user has already set up. Credentials remain on the user desktop under the control of Lens, and are not copied into AI tools. This approach reduces manual integration work and avoids passing kubeconfig files around.

Built for the tools developers already use

Lens Desktop enables Claude Code, ChatGPT, Cursor, Git Hub Copilot, and other MCP compatible clients to discover and connect to Kubernetes clusters without custom plug ins or manual integration work. Teams can use their preferred AI tools while relying on Lens to handle cluster discovery, context, and secure connectivity. The setup is designed to be straightforward. Install or update Lens Desktop, keep working with your existing AI assistant, and let Lens provide the cluster bridge when the assistant requests it.

Extending AI assisted operations with Lens Prism

The MCP capability builds on Lens Prism, the built in AI assistant for Kubernetes troubleshooting in Lens Desktop. Prism helps interpret cluster signals and suggest next steps for diagnosis and remediation. With MCP support, Lens extends beyond its own assistant and shares operational context with a broader ecosystem of AI tools. This lets teams combine their favorite coding assistants with the runtime awareness that comes from Lens, connecting code level guidance with actionable insights about live clusters.

Unified access across on premises and cloud environments

Lens helps developers and platform engineers discover, connect to, and manage Kubernetes clusters across on premises and cloud platforms with built in integrations for AWS and Azure, with Google Cloud support coming soon. The MCP server inherits this reach. When an assistant asks Lens for clusters, it can surface resources from multiple environments through one consistent interface. This unified view reduces context switching and makes it easier to move from code to deployment to operations.

Practical scenarios that benefit teams

With the MCP server, a coding assistant can locate the right cluster for a service, check pod health, view logs, describe deployments, or suggest kubectl commands, all while Lens enforces the user access model. During a production incident, an assistant can ask Lens for the current namespace context, request events and resource descriptions, and propose a rollback or scaling change for review. In a development setting, an assistant can create a new namespace, apply a manifest, and verify status without the developer leaving their editor. By centralizing connectivity in Lens, each of these workflows becomes more predictable and reduces the chance of misconfiguring access in multiple tools.

Security and governance considerations

Security and control are core to this design. Lens uses the same authentication and authorization that users already rely on for cluster access. Credentials and tokens stay local on the desktop where Lens is running. The MCP server exposes only what is needed for assistants to operate in the user context, respecting least privilege and existing role based access controls. Centralizing the bridge in Lens also makes it easier for platform teams to standardize how assistants connect, rather than approving a proliferation of custom integrations.

What this means for developers and platform engineers

The addition of MCP support to Lens shifts AI from being a separate coding helper to becoming an integrated part of day to day Kubernetes operations. Developers can move faster from code intent to cluster actions, while platform engineers can guide best practices and standardize connectivity through one trusted tool. This alignment helps teams reduce manual effort, streamline onboarding for new projects, and keep operational knowledge close to where the work happens.

Availability and getting started

The built in Lens MCP server is available in the latest release of Lens Desktop. Users can download or update from lenshq.io, then connect their preferred MCP compatible assistant. There is no need to install custom plug ins for each tool. Once Lens is managing your clusters, assistants that support MCP can request discovery and context as needed. The result is a simpler, more consistent way to let AI work with Kubernetes while keeping access and credentials under your control.