gsoc: add kubeflow sdk mcp as gsoc 2026 project idea by jaiakash · Pull Request #4290 · kubeflow/website

jaiakash · 2026-01-26T14:24:05Z

This PR proposes "MCP for Kubeflow SDK: as project idea for GSoC 2026.

Related Issue: kubeflow/sdk#238

jaiakash · 2026-01-26T14:25:21Z

/area gsoc
/area llm
/area sdk

google-oss-prow · 2026-01-26T14:25:26Z

@jaiakash: The label(s) area/llm, area/sdk cannot be applied, because the repository doesn't have them.

Details

In response to this:

/area gsoc
/area llm
/area sdk

Instructions for interacting with me using PR comments are available here. If you have questions or suggestions related to my behavior, please file an issue against the kubernetes/test-infra repository.

Copilot

Pull request overview

This PR proposes adding "MCP Server for Kubeflow SDK" as a project idea for Google Summer of Code 2026. The project aims to create a Model Context Protocol (MCP) server to enhance developer experience by enabling LLM-powered interaction with Kubeflow SDK training jobs, including status monitoring, debugging, and natural language interaction capabilities.

Changes:

Added a new GSoC 2026 project proposal for implementing an MCP server for Kubeflow SDK
Defined project scope including training log exposure, job management CRUD operations, LLM-enhanced validation, and telemetry analysis
Specified project difficulty as Medium with either 175 or 350 hours commitment

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

content/en/events/upcoming-events/gsoc-2026.md

andreyvelich

/cc @akshaychitneni @kubeflow/kubeflow-sdk-team @kubeflow/kubeflow-trainer-team

google-oss-prow · 2026-01-26T16:04:51Z

@andreyvelich: GitHub didn't allow me to request PR reviews from the following users: kubeflow/kubeflow-sdk-team, kubeflow/kubeflow-trainer-team.

Note that only kubeflow members and repo collaborators can review this PR, and authors cannot review their own PRs.

Details

In response to this:

/cc @akshaychitneni @kubeflow/kubeflow-sdk-team @kubeflow/kubeflow-trainer-team

Instructions for interacting with me using PR comments are available here. If you have questions or suggestions related to my behavior, please file an issue against the kubernetes/test-infra repository.

andreyvelich · 2026-01-26T16:04:57Z

cc @franciscojavierarceo

dhanishaphadate · 2026-01-26T16:22:24Z

@jaiakash I would like to co-mentor this project with you.

abhijeet-dhumal · 2026-01-29T14:27:19Z

Hi @jaiakash , May I opt for co-mentoring on these efforts ?

andreyvelich · 2026-01-29T23:12:25Z

content/en/events/upcoming-events/gsoc-2026.md

+
+**Components:** [kubeflow/sdk](https://www.github.com/kubeflow/sdk), [kubeflow/trainer](https://www.github.com/kubeflow/trainer)
+
+**Mentors:** [TBD], [@jaiakash](https://www.github.com/jaiakash)


Please add @dhanishaphadate and @abhijeet-dhumal as co-mentors.

Hi @dhanishaphadate @abhijeet-dhumal thanks for volunteering, added you both as co mentor. Looking forward for a successful journey ahead.

andreyvelich · 2026-01-29T23:12:46Z

content/en/events/upcoming-events/gsoc-2026.md

+
+Most of us do use LLMs to create/debug code for jobs, models, etc., but as of now there is no mechanism for the LLM to see the `TrainJob` status, debug a crash loop or provide consolidate metric about our previous task. We want to extend and improve the Developer Experience (DX) with the introduction of a **Model Context Protocol (MCP)** server for Kubeflow ecosystem.
+
+This project aims to initiate the Kubeflow MCP project with first-principles features. The MCP server will help with training jobs, their lifecycle, and associated observability data, enabling natural-language interaction through MCP-compatible clients. Tracking issue: [kubeflow/sdk#238](https://github.com/kubeflow/sdk/issues/238)


you need to update the scope for this project.

Hi @abhijeet-dhumal @dhanishaphadate please review the scope, as of now i have kept it open ended. We need some good ideas for initial draft before merging.

We REALLY need to merge it asap, but exactly defined scope and features we can always update during next phase of GSoC timeline.

Thanks for the feedback @andreyvelich! Here's a suggested scope based on the existing MVP and KEP-936:

Core Deliverables (Required):

MCP tools for TrainJob lifecycle: fine_tune(), get_training_job(), list_training_jobs(), delete_training_job()

Pre-flight validation: get_cluster_resources(), estimate_resources(), check_training_prerequisites()

Job observability: get_training_logs(), get_job_events()

Storage setup: setup_training_storage() (PVC creation for checkpoints)

Stretch Goals:

Policy-based access control (persona-based RBAC)

Custom trainer support via run_custom_training() and run_container_job()

Integration with Model Registry MCP catalog

Checkpoint listing (requires spawning helper pods)

Progress tracking (KEp-2779)

@jaiakash let me know if this aligns with your vision for the project scope.

Looks good to me, will add in PR.

Thanks @jaiakash 🙌

Look good to me as well.

Job observability: get_training_logs(), get_job_events()

As we have added this , I can scope #4294 project.

abhijeet-dhumal · 2026-01-30T05:47:19Z

content/en/events/upcoming-events/gsoc-2026.md

+
+**Contributor:**
+
+**Details:**


Small suggestion for the Details section - we could add more context about what already exists:

We already have a KEP for this project. The contributor will extend the MCP server to cover additional use cases, improve error handling, add comprehensive documentation, and potentially integrate with other Kubeflow components.

This sets clearer expectations that contributors will be building on existing work rather than starting from scratch.

but not a hard opinion, this can be skipped if not needed

Yeah, can you share a draft for that. Would be helpful.

Is this draft looks good? :

Details: The Kubeflow SDK allows users with limited Kubernetes knowledge to use standard Python APIs to interact with the Kubeflow ecosystem. Documentation(https://sdk.kubeflow.org/en/latest/index.html) Most of us use LLMs to create/debug code for jobs, models, etc., but currently there is no mechanism for the LLM to see TrainJob status, debug a crash loop, or provide consolidated metrics about previous tasks. We want to extend and improve the Developer Experience (DX) with a Model Context Protocol (MCP) server for the Kubeflow ecosystem. We have a KEP(https://github.com/kubeflow/community/issues/936) and existing MVP for this project. The contributor will extend the MCP server to cover additional use cases, improve error handling, add comprehensive documentation, and potentially integrate with other Kubeflow components like Model Registry. Core Deliverables: - MCP tools for TrainJob lifecycle (fine_tune, get_training_job, list_training_jobs, delete_training_job) - Pre-flight validation (get_cluster_resources, estimate_resources, check_training_prerequisites) - Job observability (get_training_logs, get_job_events) - Storage setup (setup_training_storage) Stretch Goals: - Policy-based access control (persona-based RBAC) - Custom trainer support (run_custom_training, run_container_job) - Integration with Model Registry MCP catalog - Progress tracking (pending KEP-2779) Tracking issue: https://github.com/kubeflow/sdk/issues/238

Looks good to me, @dhanishaphadate what about you?

Yes, adding details about what currently exists is a good idea. Once the initial work is complete and the foundation is in place, we can add those details later when the project is published on the website.

abhijeet-dhumal · 2026-01-30T05:48:30Z

Thank you @jaiakash and @dhanishaphadate !! Excited to work on this together! 🎉

jaiakash · 2026-01-30T15:58:00Z

cc @andreyvelich @abhijeet-dhumal @dhanishaphadate
PTAL We can merge this now.

Copilot

Pull request overview

Copilot reviewed 1 out of 1 changed files in this pull request and generated 5 comments.

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

Copilot · 2026-01-30T16:01:23Z

content/en/events/upcoming-events/gsoc-2026.md


+### Project 2: MCP Server for Kubeflow SDK
+
+**Components:** [kubeflow/sdk](https://www.github.com/kubeflow/sdk), [kubeflow/trainer](https://www.github.com/kubeflow/trainer)


The URLs use https://www.github.com/ but the majority of GitHub links in this file use https://github.com/ without the "www" subdomain (e.g., lines 79, 116, 146, 159, 161, 178, 192, 194). For consistency with the established convention in this file, remove "www." from these URLs.

Copilot · 2026-01-30T16:01:24Z

content/en/events/upcoming-events/gsoc-2026.md

+
+**Components:** [kubeflow/sdk](https://www.github.com/kubeflow/sdk), [kubeflow/trainer](https://www.github.com/kubeflow/trainer)
+
+**Mentors:** [@jaiakash](https://www.github.com/jaiakash), [@dhanishaphadate](https://www.github.com/dhanishaphadate), [@abhijeet-dhumal](https://www.github.com/abhijeet-dhumal)


The mentor GitHub URLs use https://www.github.com/ but should use https://github.com/ without the "www" subdomain for consistency with other mentor links (see line 178 in Project 3).

Copilot · 2026-01-30T16:01:24Z

content/en/events/upcoming-events/gsoc-2026.md

+
+Most of us use LLMs to create/debug code for jobs, models, etc., but currently there is no mechanism for the LLM to see TrainJob status, debug a crash loop, or provide consolidated metrics about previous tasks. We want to extend and improve the Developer Experience (DX) with a Model Context Protocol (MCP) server for the Kubeflow ecosystem.
+
+We have a [kubeflow/community#936](https://github.com/kubeflow/community/issues/936) and existing MVP for this project. The contributor will extend the MCP server to cover additional use cases, improve error handling, add comprehensive documentation, and potentially integrate with other Kubeflow components like Model Registry.


Missing article before "existing". The sentence should read "We have a [kubeflow/community#936]...and an existing MVP for this project."

Suggested change

We have a [kubeflow/community#936](https://github.com/kubeflow/community/issues/936) and existing MVP for this project. The contributor will extend the MCP server to cover additional use cases, improve error handling, add comprehensive documentation, and potentially integrate with other Kubeflow components like Model Registry.

We have a [kubeflow/community#936](https://github.com/kubeflow/community/issues/936) and an existing MVP for this project. The contributor will extend the MCP server to cover additional use cases, improve error handling, add comprehensive documentation, and potentially integrate with other Kubeflow components like Model Registry.

Copilot · 2026-01-30T16:01:24Z

content/en/events/upcoming-events/gsoc-2026.md

+* Experience with LLM / MCP development.
+* Familiarity with the Kubeflow SDK and Trainer codebase.
+* Understanding of the Kubeflow Ecosystem and basic Kubernetes concepts.
+* Engage and contribute to Kubeflow community on Slack and GitHub.


The bullet point style uses asterisks (*) while Project 1 uses hyphens (-) for its "Skills Required/Preferred" list (lines 126-130). For consistency within the document, consider using hyphens instead of asterisks.

Copilot · 2026-01-30T16:01:24Z

content/en/events/upcoming-events/gsoc-2026.md

+* Familiarity with the Kubeflow SDK and Trainer codebase.
+* Understanding of the Kubeflow Ecosystem and basic Kubernetes concepts.
+* Engage and contribute to Kubeflow community on Slack and GitHub.
 ---


This horizontal rule separator is inconsistent with the document structure. Project 1 does not have a separator after it, and Project 3 presumably doesn't either. This separator should be removed to maintain consistency with the other project entries.

abhijeet-dhumal · 2026-01-30T18:17:55Z

/lgtm
Nit pick : Some of the co-pilot reviews seems legit, wdyt @jaiakash ? 👀

juliusvonkohout · 2026-02-01T14:45:58Z

Please rebase after #4286. the deadline is close @andreyvelich

Signed-off-by: Akash Jaiswal <akashjaiswal3846@gmail.com>

jaiakash · 2026-02-02T08:06:37Z

Please rebase after #4286. the deadline is close @andreyvelich

Done, Please review @andreyvelich @abhijeet-dhumal @dhanishaphadate

Copilot

Pull request overview

Copilot reviewed 1 out of 1 changed files in this pull request and generated 2 comments.

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

Copilot · 2026-02-02T08:38:12Z

content/en/events/upcoming-events/gsoc-2026.md

+
+Most of us use LLMs to create/debug code for jobs, models, etc., but currently there is no mechanism for the LLM to see TrainJob status, debug a crash loop, or provide consolidated metrics about previous tasks. We want to extend and improve the Developer Experience (DX) with a Model Context Protocol (MCP) server for the Kubeflow ecosystem.
+
+We have a [kubeflow/community#936](https://github.com/kubeflow/community/issues/936) and an existing MVP for this project. The contributor will extend the MCP server to cover additional use cases, improve error handling, add comprehensive documentation, and potentially integrate with other Kubeflow components like Model Registry.


The reference format is inconsistent with markdown conventions. The link text "kubeflow/community#936" should be a descriptive text, not a repository reference notation. Consider changing this to match the pattern used elsewhere in the document, such as: "We have an open issue tracking this work and an existing MVP for this project."

Suggested change

We have a [kubeflow/community#936](https://github.com/kubeflow/community/issues/936) and an existing MVP for this project. The contributor will extend the MCP server to cover additional use cases, improve error handling, add comprehensive documentation, and potentially integrate with other Kubeflow components like Model Registry.

We have an [open tracking issue](https://github.com/kubeflow/community/issues/936) and an existing MVP for this project. The contributor will extend the MCP server to cover additional use cases, improve error handling, add comprehensive documentation, and potentially integrate with other Kubeflow components like Model Registry.

Copilot · 2026-02-02T08:38:12Z

content/en/events/upcoming-events/gsoc-2026.md

+- Integration with Model Registry MCP catalog
+- Progress tracking (pending [KEP-937](https://github.com/kubeflow/community/pull/937))
+
+Tracking issue: https://github.com/kubeflow/sdk/issues/238


The tracking issue link format is inconsistent with other projects in this file. Project 2 uses the format [kubeflow/katib#2605](https://github.com/kubeflow/katib/issues/2605) (line 152), but this project uses a plain URL. For consistency, consider formatting this as: Tracking issue: [kubeflow/sdk#238](https://github.com/kubeflow/sdk/issues/238)

andreyvelich

/lgtm
/approve

google-oss-prow · 2026-02-02T12:11:51Z

[APPROVALNOTIFIER] This PR is APPROVED

This pull-request has been approved by: andreyvelich

The full list of commands accepted by this bot can be found here.

The pull request process is described here

Details

Needs approval from an approver in each of these files:

~~content/en/events/OWNERS~~ [andreyvelich]

Approvers can indicate their approval by writing /approve in a comment
Approvers can cancel approval by writing /approve cancel in a comment

Signed-off-by: Akash Jaiswal <akashjaiswal3846@gmail.com>

Copilot AI review requested due to automatic review settings January 26, 2026 14:24

google-oss-prow bot added the area/community AREA: Community Docs label Jan 26, 2026

google-oss-prow bot requested review from jbottum and thesuperzapper January 26, 2026 14:24

google-oss-prow bot added the size/M label Jan 26, 2026

Copilot started reviewing on behalf of jaiakash January 26, 2026 14:24 View session

google-oss-prow bot added the area/gsoc AREA: Google Summer of Code label Jan 26, 2026

Copilot AI reviewed Jan 26, 2026

View reviewed changes

andreyvelich reviewed Jan 26, 2026

View reviewed changes

google-oss-prow bot requested a review from akshaychitneni January 26, 2026 16:04

jaiakash force-pushed the gsoc26-sdk-mcp branch from a959cea to 93c2674 Compare January 26, 2026 16:23

jaiakash mentioned this pull request Jan 27, 2026

Built an MCP Server for Kubeflow SDK kubeflow/sdk#238

Open

andreyvelich reviewed Jan 29, 2026

View reviewed changes

jaiakash force-pushed the gsoc26-sdk-mcp branch from d5926fd to 6e9c363 Compare January 30, 2026 03:47

abhijeet-dhumal reviewed Jan 30, 2026

View reviewed changes

jaiakash requested a review from Copilot January 30, 2026 15:58

Copilot started reviewing on behalf of jaiakash January 30, 2026 15:58 View session

Copilot AI reviewed Jan 30, 2026

View reviewed changes

google-oss-prow bot assigned abhijeet-dhumal Jan 30, 2026

google-oss-prow bot added the lgtm label Jan 30, 2026

gsoc: add kubeflow sdk mcp as gsoc 2026 project idea

a50f3ac

Signed-off-by: Akash Jaiswal <akashjaiswal3846@gmail.com>

jaiakash force-pushed the gsoc26-sdk-mcp branch from ef7299d to a50f3ac Compare February 2, 2026 08:05

google-oss-prow bot removed the lgtm label Feb 2, 2026

jaiakash requested review from abhijeet-dhumal, Copilot and dhanishaphadate February 2, 2026 08:35

Copilot started reviewing on behalf of jaiakash February 2, 2026 08:35 View session

Copilot AI reviewed Feb 2, 2026

View reviewed changes

andreyvelich reviewed Feb 2, 2026

View reviewed changes

google-oss-prow bot assigned andreyvelich Feb 2, 2026

google-oss-prow bot added the lgtm label Feb 2, 2026

google-oss-prow bot added the approved label Feb 2, 2026

google-oss-prow bot merged commit 168a82a into kubeflow:master Feb 2, 2026
12 of 13 checks passed

ederign mentioned this pull request Feb 2, 2026

gsoc: Add Composable Kale Notebooks with Visual Pipeline Editor proposal #4295

Merged

Shekharrajak pushed a commit to Shekharrajak/website that referenced this pull request Feb 2, 2026

gsoc: add kubeflow sdk mcp as gsoc 2026 project idea (kubeflow#4290)

8f29582

Signed-off-by: Akash Jaiswal <akashjaiswal3846@gmail.com>

ederign pushed a commit to ederign/website that referenced this pull request Feb 2, 2026

gsoc: add kubeflow sdk mcp as gsoc 2026 project idea (kubeflow#4290)

e24d4da

Signed-off-by: Akash Jaiswal <akashjaiswal3846@gmail.com>


		Components: [kubeflow/sdk](https://www.github.com/kubeflow/sdk), [kubeflow/trainer](https://www.github.com/kubeflow/trainer)

		Mentors: [TBD], [@jaiakash](https://www.github.com/jaiakash)


		Most of us do use LLMs to create/debug code for jobs, models, etc., but as of now there is no mechanism for the LLM to see the `TrainJob` status, debug a crash loop or provide consolidate metric about our previous task. We want to extend and improve the Developer Experience (DX) with the introduction of a Model Context Protocol (MCP) server for Kubeflow ecosystem.

		This project aims to initiate the Kubeflow MCP project with first-principles features. The MCP server will help with training jobs, their lifecycle, and associated observability data, enabling natural-language interaction through MCP-compatible clients. Tracking issue: [kubeflow/sdk#238](https://github.com/kubeflow/sdk/issues/238)


		### Project 2: MCP Server for Kubeflow SDK

		Components: [kubeflow/sdk](https://www.github.com/kubeflow/sdk), [kubeflow/trainer](https://www.github.com/kubeflow/trainer)


		Components: [kubeflow/sdk](https://www.github.com/kubeflow/sdk), [kubeflow/trainer](https://www.github.com/kubeflow/trainer)

		Mentors: [@jaiakash](https://www.github.com/jaiakash), [@dhanishaphadate](https://www.github.com/dhanishaphadate), [@abhijeet-dhumal](https://www.github.com/abhijeet-dhumal)


		Most of us use LLMs to create/debug code for jobs, models, etc., but currently there is no mechanism for the LLM to see TrainJob status, debug a crash loop, or provide consolidated metrics about previous tasks. We want to extend and improve the Developer Experience (DX) with a Model Context Protocol (MCP) server for the Kubeflow ecosystem.

		We have a [kubeflow/community#936](https://github.com/kubeflow/community/issues/936) and existing MVP for this project. The contributor will extend the MCP server to cover additional use cases, improve error handling, add comprehensive documentation, and potentially integrate with other Kubeflow components like Model Registry.

	We have a [kubeflow/community#936](https://github.com/kubeflow/community/issues/936) and an existing MVP for this project. The contributor will extend the MCP server to cover additional use cases, improve error handling, add comprehensive documentation, and potentially integrate with other Kubeflow components like Model Registry.
	We have an [open tracking issue](https://github.com/kubeflow/community/issues/936) and an existing MVP for this project. The contributor will extend the MCP server to cover additional use cases, improve error handling, add comprehensive documentation, and potentially integrate with other Kubeflow components like Model Registry.


		Contributor:

		Details:

Conversation

jaiakash commented Jan 26, 2026

Uh oh!

jaiakash commented Jan 26, 2026

Uh oh!

google-oss-prow bot commented Jan 26, 2026

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

andreyvelich left a comment

Choose a reason for hiding this comment

Uh oh!

google-oss-prow bot commented Jan 26, 2026

Uh oh!

andreyvelich commented Jan 26, 2026

Uh oh!

dhanishaphadate commented Jan 26, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

abhijeet-dhumal commented Jan 29, 2026

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

abhijeet-dhumal commented Jan 30, 2026

Uh oh!

jaiakash commented Jan 30, 2026

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Uh oh!

Copilot AI Jan 30, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Jan 30, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Jan 30, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Jan 30, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Jan 30, 2026

Choose a reason for hiding this comment

Uh oh!

abhijeet-dhumal commented Jan 30, 2026

Uh oh!

dhanishaphadate commented Jan 26, 2026 •

edited

Loading