Adopt MCP servers at scale with Kubernetes
Tackle the full MCP lifecycle with infrastructure you trust
What can you do with Stacklok’s MCP runtime?
We’re so glad you asked! You can:
- Run MCP servers as native Kubernetes resources
- Automate deployment, proxying and lifecycle management of MCP servers
- Configure rate limits on MCP server resources to prevent noisy neighbors and protect downstream services
- Define security guardrails like RBAC, least privilege service accounts, restricted network access and more
- Set up a token exchange so MCP servers can authenticate to backend services
- Work with the co-creators of Kubernetes (Craig McLuckie and Joe Beda) to get the most from your K8s investments
Let’s get a little more technical. Here’s what separates Stacklok’s MCP runtime from alternatives.
Completeness
Alternatives: Most gateways and control planes don’t include a runtime at all; they are just a policy and visibility layers.
Stacklok: We felt it was imperative that our platform have an integrated runtime. We understand that enterprises need to manage the MCP lifecycle with permission profiles, network isolation and more.
Containerized
Alternatives: Many MCP solutions either don’t leverage containers or wrap MCP servers in “generic Docker” which weakens security.
Stacklok: Stacklok dynamically builds a container image for each MCP server, so you benefit from isolation and can enforce identity and access policy per request.
Centralized
Alternatives: Most solutions require that your platform team build its own structure to manage a growing fleet of MCP servers.
Stacklok: Stacklok’s runtime includes a registry server you can use for cluster-wide namespace scanning across multi-tenant Kubernetes deployments.
Admin-friendly
Alternatives: Other solutions built on Kubernetes use CRDs that the platform team understands, but that don’t scale to admins.
Stacklok: We make it easy for platform teams to use Kubernetes patterns like CRDs, but also offer a database-backed system, so that admins can manage servers, skills and models without dependencies.
Fun fact:
Stacklok’s Kubernetes operator enables 10M+ tool calls every month.
Want to talk through your options?