mirobody

Why we included this project

Mirobody is a self-hosted health data engine that links personal medical records to a language model while keeping everything on your own device or server. It pulls wearable, lab, and clinical-EHR data from a long list of sources, including aggregators like Terra and Human API, device brands like Fitbit and Garmin, and SMART-on-FHIR EHRs such as Epic, then normalizes everything to FHIR R4 with UCUM unit handling and SNOMED/LOINC mapping before exposing it to LLMs through one interface. A Model Context Protocol endpoint lets clients like Claude Desktop or Cursor discover and call health tools directly, so you can ask plain-language questions about trends and get answers grounded in your actual records. The native apps can also run a small Gemma or Qwen model fully on-device, with no server and no network needed, which matters for privacy-sensitive health work. That combination makes it worth a look for developers and small teams building secure health assistants that need controlled access to personal data.