If you have used Claude Code, Cursor, or OpenAI Codex, you have already experienced Agentic AI, whether it is coding an application, creating a quick script, or debugging an issue. The use cases are infinite, and like many, you can easily burn through your token quota when using the latest state-of-the-art Frontier AI models, which are simply amazing!
While spending some time playing with our latest VCF Private AI Services (PAIS) release (here and here), I have also been trading notes with colleagues doing something similar but rolling their own local AI stack using just vSphere Kubernetes Service (VKS) running on VMware Cloud Foundation (VCF) 9.1.1. Through those conversations, I came to learn about Pi, not the number, but a lightweight coding agent and agent harness that can connect to a number of AI model providers, including those providing an OpenAI-compatible endpoint.
Since VCF PAIS provides an OpenAI-compatible endpoint that can serve one or more AI models, you can now have your own Agentic AI using Pi running within your on-premises environment using various open weight models! 😎
