Field notes on Cloud Foundry, Tanzu, BOSH, GenAI tiles, GPU homelabs, and whatever else is currently humming in the rack — written from hands-on builds, lab notes, and the troubleshooting that turns docs into working systems. By Nick Kuhn.
The model call can stay with Anthropic, Bedrock, or Vertex. The agent harness is the part you get to place. Here's why Tanzu Platform is a natural runtime for Claude agents.
Push claude remote-control to Cloud Foundry and drive a Claude agent from the mobile app while it executes inside your foundation — the GitHub self-hosted-runner model, applied to AI.
Wrap the Claude Agent SDK in a small Cloud Foundry service and call it from your apps and pipelines — a code reviewer, an incident triager, a chat UI — with the worker running near your internal systems.
Build a zero-code research assistant on Tanzu Platform 10.4 with AGENTS.md, the agent buildpack, AI Services, and cf push.
Stand up the Tanzu Platform 10.4 MCP gateway, federate two MCP servers behind it, and bind each gateway path to your AI agent as a tagged MCP service.
Use the Tanzu MCP gateway as a developer landing zone for Cursor, Claude Code, Goose, and Codex, with platform-auth pass-through for web search and per-user GitHub OAuth.
A complete recipe for adding Gemma 4 31B (multimodal) as a vLLM model on the Tanzu AI Services tile — fields, the single-GPU Blackwell setup, and the performance numbers you'll see.
A complete recipe for adding GPT-OSS-120B as a vLLM model on the Tanzu AI Services tile — fields, MoE tunings, and the performance numbers you'll see on a 4× RTX 3090 worker.
A complete recipe for adding Qwen3.5-27B as a vLLM model on the Tanzu AI Services tile — fields, chat template, tunings, and the performance numbers you'll see.
A platform-engineer's recipe for enabling block storage on Tanzu Platform 10.4 with the Elastic Application Runtime tile — unlocking stateful workloads that need persistent disk and were previously hard to land on the platform.
Comprehensive guide for navigating the migration from TanzuNet to Broadcom's Customer Support Portal (BSP).
Building a portable 8-liter homelab for the VMware Explore Barcelona Hackathon featuring GenAI for TAS.
A curated guide for TAS-focused sessions and activities at VMware Explore Barcelona.
Team TAS takes first place with GenAI for Tanzu Application Service, enabling LLMs and GPU-equipped VMs.
Analysis of homelab power consumption using smart plugs — approximately 21 kWh daily at ~$1,226 annually.
Curated recommendations for TAS-related sessions featuring platform engineering demos and customer panels.
Comprehensive documentation of building a 7-host homelab using HP z440 workstations with vSAN and 10Gb networking.