Open source · Self-hosted · Apache 2.0
Turn scattered health data into answers you can trace.
Lab reports, wearables and genetic files in one standardized record. Start with Docker and one model key, or keep every model on your own machine.
Three reports, three names for the same test. One trend, with the file behind every number.
What is Mirobody?
Mirobody turns scattered health data into answers you can trace. It reads lab reports, wearable data, photos and genetic files, gives each value one standard code and unit, and lets an agent answer questions over the record while naming the file behind every number. The engine is open source under Apache 2.0, built by Theta Health, and runs under its Theta Wellness app.
Collect, translate, and ask: the three stages from a source file to a traceable answer.
Start from what you need
Run it with Docker
The complete web app with a demo record:
./deploy.sh, then one model key or local models. Records stay in a Postgres you run.See it in five scenes
Upload, standardize, ask across reports, share with family, write a journal entry.
Keep every model on your machine
No model key: llama.cpp serves the models beside the stack. What each size needs, per platform.
Connect your own agent
Your record in Claude Code, Codex, Cursor or Gemini CLI over MCP.
Resolve names in Python
pip install mirobody: names to LOINC and units to UCUM, offline, with no key.Deploy on a server
Production settings, secrets, HTTPS, upgrades and backups.