# K8s Ai

> 🧠 Tenant repository bootstrapped by k8s-infrastructure that contains the manifests for AI related applications

- **Type:** MCP server
- **Install:** `agentstack add mcp-mmontes11-k8s-ai`
- **Verified:** Yes — security-reviewed for prompt injection and unsafe behavior
- **Seller:** [mmontes11](https://agentstack.voostack.com/s/mmontes11)
- **Installs:** 0
- **Category:** [Cloud & Infrastructure](https://agentstack.voostack.com/c/cloud-infrastructure)
- **Latest version:** 0.1.0
- **License:** Apache-2.0
- **Upstream author:** [mmontes11](https://github.com/mmontes11)
- **Source:** https://github.com/mmontes11/k8s-ai

## Install

```sh
agentstack add mcp-mmontes11-k8s-ai
```

Requires the [AgentStack CLI](https://agentstack.voostack.com/docs/cli). Works with Claude Code, Cursor, and any MCP-compatible agent.

## About

# 🧠 k8s-ai

Tenant repository bootstrapped by [k8s-infrastructure](https://github.com/mmontes11/k8s-infrastructure) that contains the manifests for AI related applications

## Overview

This repository manages AI workloads on Kubernetes using GitOps with Flux CD. It includes deployments for LLM inference services, web UIs, and model serving infrastructure.

## Applications

### Open WebUI
- **Path**: `./apps/open-webui`
- **Type**: HelmRelease (ollama-webui chart)
- **Description**: Web interface for interacting with LLMs
- **Features**: 
  - Persistent storage via PVC
  - Integration with Ollama backend
  - Model access control bypass enabled

### ComfyUI
- **Path**: `./apps/comfyui`
- **Type**: Native Kubernetes resources
- **Description**: Graph-based interface for Stable Diffusion
- **Image**: [mmontes11/docker-comfyui](https://github.com/mmontes11/docker-comfyui)
- **Features**:
  - Persistent volume for model caching
  - Replication source/destination for data synchronization
  - RESTic backup support

### n8n
- **Path**: `./apps/n8n`
- **Type**: HelmRelease (n8n helm chart)
- **Description**: Workflow automation and integration platform
- **Features**:
  - Persistent storage via PVC
  - RESTic backup support
  - Replication source/destination for data synchronization

### opencode
- **Path**: `./apps/opencode`
- **Type**: Native Kubernetes resources
- **Description**: Coding agent and AI workspace for interactive development
- **Image**: [mmontes11/docker-opencode](https://github.com/mmontes11/docker-opencode)
- **Features**:
  - NVIDIA GPU support for accelerated model training and inference
  - Persistent storage (100Gi PVC)
  - Pre-configured development environment with tools
  - RESTic backup support
  - Replication source/destination for data synchronization
  - Integration with GitHub, HuggingFace, and n8n via tokens

## Infrastructure

### Model Serving

#### Ollama
- **Path**: `./infrastructure/ollama`
- **Description**: Lightweight LLM inference server
- **Features**:
  - Native GPU support
  - Simple HTTP API
  - Model caching

#### llama.cpp
- **Path**: `./infrastructure/llamacpp`
- **Description**: High-performance C/C++ inference engine optimized for CPU and GPU
- **Features**:
  - Qwen3.6 MTP model support with 1.4-2.2x faster inference
  - 256k context window for agentic AI workflows
  - StatefulSet deployment with persistent storage
  - Prometheus ServiceMonitor integration
  - Ingress routing via HTTPRoute

#### vLLM
- **Path**: `./infrastructure/vllm`
- **Description**: High-throughput LLM serving with PagedAttention
- **Use Case**: Production workloads requiring high concurrency

#### KServe
- **Path**: `./infrastructure/kserve`
- **Example**: `./examples/llminferenceservice.yaml`
- **Description**: Kubernetes-native ML serving platform
- **Features**:
  - LLMInferenceService CRD
  - Custom model templates

### MCP Servers
- **MCP Kubernetes**: Kubernetes model context protocol server
- **MCP Grafana**: Grafana monitoring integration
- **MCP GitHub**: GitHub API integration
- **MCP Photoprism**: Photo management (mmontes & xiaowen)

## Architecture

```
├── apps/                    # Application deployments
│   ├── comfyui/            # ComfyUI deployment
│   ├── n8n/                # n8n workflow automation
│   ├── opencode/           # opencode AI development workspace
│   └── open-webui/         # Open WebUI deployment
├── clusters/               # Cluster-specific configurations
│   └── homelab/
│       ├── apps.yaml       # Application Kustomizations
│       ├── infrastructure.yaml
│       └── namespaces.yaml
├── examples/               # Example configurations
│   └── llminferenceservice.yaml
└── infrastructure/         # Shared infrastructure
    ├── kserve/             # KServe ML serving
    ├── vllm/               # vLLM serving engine
    ├── llamacpp/           # llama.cpp inference engine
    ├── lws/                # LeaderWorkerSet
    ├── ollama/             # Ollama LLM backend
    └── mcp-*/             # MCP server integrations
```

## AI Benchmarks

LLM benchmarks using llama.cpp on Kubernetes: [mmontes11/llm-bench](https://github.com/mmontes11/llm-bench)

## License

[MIT](./LICENSE)

## Source & license

This open-source MCP server is cataloged on AgentStack and links to its original source — we do not rehost the code.

- **Author:** [mmontes11](https://github.com/mmontes11)
- **Source:** [mmontes11/k8s-ai](https://github.com/mmontes11/k8s-ai)
- **License:** Apache-2.0

Install and usage instructions live in the source repository linked above.

## Pricing

- **Free** — Free

## Security capabilities

Automated source analysis of v0.1.0 — what this tool can access:

- **Network access:** no
- **Filesystem access:** no
- **Shell / process execution:** no
- **Environment & secrets:** no
- **Dynamic code execution:** no

*"Yes" means the capability is present in the source — more access means more to trust, not that it is unsafe.*


## Versions

- **0.1.0** — security scan: passed — Imported from the upstream source.

## Links

- Listing page: https://agentstack.voostack.com/l/mcp-mmontes11-k8s-ai
- Seller: https://agentstack.voostack.com/s/mmontes11
- Browse the marketplace: https://agentstack.voostack.com/browse

---
Listed on AgentStack — the marketplace for AI agent skills and MCP servers. Every listing is security-reviewed. Creators keep 70%.