The first iteration of Omni-Endo AI showed me that handing Glooko data to an LLM could produce some interesting ideas and give some great insights. But it required a manual copy-and-paste handoff. This was done to avoid LLM tie-in and also to avoid API costs.
This new version completely removes that friction by rebuilding the tool around the Model Context Protocol (MCP). Think of MCP as a secure extension for your AI. Instead of exporting text files, the AI now reaches directly into your data to pull and analyze what it needs mid-conversation. You can simply ask:
"Why do I keep spiking after dinner?" "Show me my worst day this month and dissect what happened."
It bridges your Glooko history (bolus records, CGM trends, settings) straight into Claude, or a completely local setup using Ollama if you want to keep the data totally private.
The core philosophy hasn't changed. There are no subscription fees, no third-party API servers, and your credentials stay entirely on your own machine. The AI persona also remains intentionally sharp. It’s designed to skip the pleasantries and aggressively pinpoint where bolus timing is off, corrections are overdone, or basal rates are failing.
It runs via Docker and requires no coding knowledge to set up. It even comes with a system prompt you can copy and paste to walk you through the installation step by step.
If you don't use an Omnipod 5 but want to see how MCP works, I’ve bundled three months of my own data into the repo so you can test it immediately.
It’s an open-source clinical audit tool built out of personal curiousity. Dig in, use it for your next consultation, and let me know how it works for you.