Kubernetes and Model Context Protocol discussion

[Derek Carr]

Hi all,

There has been a lot of interest around the Model Context Protocol (MCP). We think it may be useful to develop a MCP server that integrates with Kubernetes resources as a building block for AI applications. A number of experimental projects have come up in this space that vary both in scope and capabilities, but we think having a project with closer affinity to Kubernetes for others to rally around to support this effort would be a benefit.  Our collective experience supporting our user community with consistent design principles [1],  object management techniques [2], security approaches [3] and composable architecture [4] could inform the evolution of this component.

What we would like input from SIG Arch on is the following:

- Does SIG Arch agree that an MCP server could be in scope for the Kubernetes project?

- If so, what SIGs would potentially be relevant to support its growth?  For example, SIG Arch, SIG CLI, SIG Auth, and SIG API Machinery seemed like relevant sponsors

- Gauge interest from other community members to kick-start creation of a repo, maintainers, code, target user personas, and perspectives on what is safe versus unsafe to expose.

We added an item to this week’s agenda to kick start the discussion.

Thanks,

Derek Carr and Mrunal Patel

[1] https://github.com/kubernetes/community/tree/master/sig-architecture#architecture-and-api-governance-1

[2] https://kubernetes.io/docs/concepts/overview/working-with-objects/object-management/

[3] https://kubernetes.io/docs/concepts/security/

[4] https://kubernetes.io/docs/concepts/extend-kubernetes/

[Antonio Ojea]

Is this MCP server integration proposal part of the core project (to live in the kubernetes/kubernetes repo) or an additional project (as in kubernetes/mcp or kubernetes-sigs/mcp)?

--
You received this message because you are subscribed to the Google Groups "kubernetes-sig-architecture" group.
To unsubscribe from this group and stop receiving emails from it, send an email to kubernetes-sig-arch...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/kubernetes-sig-architecture/CAHROWxT9C9%2BMsTc8vcNUHJ6NmR8%3DD19cf-ycu09N5cz0Yb7jmQ%40mail.gmail.com.

[Derek Carr]

Our working assumption had been for this to be an additional project per your example.

[Davanum Srinivas]

Derek,

+100 to a new repo for sure

Are we starting from scratch or is there a project repo already that is being proposed as seed?

Which SIG(s) are we looking at as sponsors (just sig-arch)?

The CNCF TAG Runtime is poking at what's happening across the cloud native landscape too if you haven't already seen:
https://github.com/cncf/tag-runtime/issues/212 

What are the benefits of doing this here under our SIG as opposed to a new CNCF project?  (we have a lot of strengths as you pointed out, but we are stretched thin as well).

Happy to talk more! next sig-arch call perhaps?

thanks,
Dims

To view this discussion visit https://groups.google.com/d/msgid/kubernetes-sig-architecture/CAHROWxTb8fZe98EfW4crvo8EtOriHCmRX%2BkCqEqj065aCGNi-A%40mail.gmail.com.

--

Davanum Srinivas :: https://twitter.com/dims

[Mrunal Patel]

On Tuesday, April 15, 2025 at 7:22:05 AM UTC-7 Davanum Srinivas wrote:

Derek,

+100 to a new repo for sure

Are we starting from scratch or is there a project repo already that is being proposed as seed?

We are proposing that we start from scratch in a new repo and work on it together as a community following the k8s design principles (outlined in first post).

Which SIG(s) are we looking at as sponsors (just sig-arch)?

 SIG-CLI/SIG-Auth/SIG-apimachinery will have possible intersections here. We wanted to bring this to sig-arch first

and then we can scope out and work with SIGs for their involvement.  

The CNCF TAG Runtime is poking at what's happening across the cloud native landscape too if you haven't already seen:
https://github.com/cncf/tag-runtime/issues/212 

What are the benefits of doing this here under our SIG as opposed to a new CNCF project?  (we have a lot of strengths as you pointed out, but we are stretched thin as well).

 We think it makes sense to have it under kubernetes-sigs for closer affinity to kubernetes. I think there is a lot of interest in this space so we 

should be able to gather enough interest from contributors and find maintainers. 

 

Happy to talk more! next sig-arch call perhaps?

Yes! Looking forward to this.  

Thanks,

Mrunal

[mpa...@redhat.com]

As a follow-up from our past SIG Arch discussion on Kubernetes and the Model Context Protocol, here is a link to a representative MCP server to demonstrate the concept.  

• https://github.com/manusa/kubernetes-mcp-server (example server)

Additionally, we would like to propose the creation of a new working group, WG AI Integration, to address the growing need for standardized and robust integration patterns between external AI/ML control planes and Kubernetes (for example via Model Context Protocol), as well as identifying system deployment patterns for AI integrations with or on Kubernetes clusters.

See the linked proposal for full details - https://docs.google.com/document/d/13OJvWKGKZL0V4nPpv9anitL5je92FCdQyXePcrY2N84/edit?tab=t.0

Our next step will be to take this proposal to stakeholder SIGs.

Thanks,

Mrunal & Derek

[Davanum Srinivas]

Sounds like plan Mrunal. thanks!

To view this discussion visit https://groups.google.com/d/msgid/kubernetes-sig-architecture/7253d0e1-0a3a-4a53-bd2b-eaaa05e3583en%40googlegroups.com.

[Mrunal]

Thanks for all the interest from the community. I have opened a PR to proceed with the WG AI Integration formation - https://github.com/kubernetes/community/pull/8519/files.

Thanks! We look forward to collaborating with you. 

To view this discussion visit https://groups.google.com/d/msgid/kubernetes-sig-architecture/CANw6fcH85R7w_HGOFgT_F_mbXo7Qo1PC7coaKSG%3D0k8f6oHRCQ%40mail.gmail.com.