Skip to main content

​
What is Mirobody?

Mirobody is an open-source, self-hosted data + agent engine for personal AI — built on the belief that the next frontier of AI is not knowing more about the world, but knowing more about you.
  • 🔌 HTTP Remote MCP Server — your tools instantly become callable by ChatGPT, Claude, Cursor, or any MCP client.
  • 🤖 Native agent runtime — a self-developed engine that reproduces Claude Code’s autonomous capabilities locally, plus support for Claude Agent Skills (SKILL.md).
  • 🔄 Pluggable data providers — bring your own data source via the Providers API; 300+ wearables, Apple Health, Google Health, and EHR coverage for ~90% of US records are already wired up.
  • 🏥 Health-grade standards — 400+ health indicators normalized to FHIR / LOINC / SNOMED CT / RxNorm with cross-vocabulary search.
Health, Finance & More. Mirobody is domain-agnostic: this docs site focuses on the health vertical (the engine behind Theta Wellness), but the same architecture powers finance, legal, devops, or any vertical where personal data meets AI. Drop your own modules into tools/ and skills/ to start.

Quick Start

One-command Docker deploy in under 5 minutes

Use Cases & Examples

Real conversations with synthetic health data

Provider Integration

Add a new device or data source in hours

API Reference

Pulse, Chat, and MCP protocol endpoints

​
Why Mirobody?

Forget JSON schemas and manual bindings — your Python code is the only definition required. A function in tools/ instantly works in ChatGPT (via Apps SDK), Claude / Cursor (via MCP), and the local web client. Mirobody runs simultaneously as an MCP client (to use tools) and an OAuth-enabled MCP server (to expose them).
The engine runs locally end-to-end — Postgres, Redis, agent runtime, MCP server, file pipeline. Bring your own LLM via OpenRouter or first-party keys (OpenAI, Gemini, Anthropic, Azure, Volcengine, Dashscope). No third-party data hop required.
  • DeepAgent — single-model tool orchestration, best for complex multi-step research
  • MixAgent — two-phase fusion: a capable model orchestrates tools, a cheaper model writes the response. Cost/quality balance for high-volume workloads.
  • BaselineAgent — direct LLM chat with no tools, ideal for low-latency Q&A
Drop a SKILL.md + metadata.json into skills/ and the agent auto-discovers it. The full Skill body is loaded into the agent’s context when activated — write detailed playbooks, examples, edge cases freely.
PDFs, Excel, CSV, audio (WAV/MP3/AIFF/AAC/OGG/FLAC), images (PNG/JPEG/WebP/HEIC/HEIF), genetic data. LLM-powered extraction maps free text to standard indicators automatically.
The built-in execute tool runs shell + Python in isolated E2B cloud sandboxes for data analysis, chart generation, and computation — with PostgreSQL-backed file persistence so write_file in one turn is visible to execute in the next.
One-command ./deploy.sh for Docker, pip install -e . for local Python dev, or MCP_PUBLIC_URL=https://yourdomain.com to expose your MCP server publicly for ChatGPT Apps.

​
Architecture at a glance

LayerPathNotes
Agent runtimemirobody/pub/agents/DeepAgent, MixAgent, BaselineAgent
LLM clientsmirobody/utils/llm/OpenAI, Gemini, Anthropic, Azure, Volcengine, Dashscope (HIPAA-compliant routing)
MCP servermirobody/mcp/JSON-RPC 2.0, local + HTTP remote
Chat servicemirobody/chat/Sessions, history, streaming (HTTP/WebSocket), memory
Data pipeline (Pulse)mirobody/pulse/300+ wearables, Apple/Google Health, EHR, file parser, aggregation
Health standardsmirobody/indicator/, mirobody/pulse/core/400+ indicators, FHIR/LOINC/SNOMED CT/RxNorm
Embeddingsmirobody/utils/embedding.py1024-dim, Gemini/Qwen, pgvector
Extension points are top-level directories — drop files in, they’re auto-discovered:
DirectoryPurpose
tools/Python tools → auto-exposed as MCP tools
skills/Claude Agent Skills (SKILL.md + metadata.json)
agents/Custom agent implementations
providers/Custom data providers
prompts/Jinja2 prompt templates
resources/Static resources exposed via MCP

​
Supported Providers (Health Vertical)

Garmin

Activity, sleep, heart rate, HRV

Whoop

Recovery, sleep, workouts, cycles

Oura

Sleep, readiness, activity

Renpho

Body composition, weight

Apple Health

iOS Health + CDA medical records

Google Health

Android health data
300+ devices are reachable via the Vital integration; EHR coverage spans ~90% of US records. Custom providers can be added in hours — see Provider Integration.

​
Tech Stack

  • Python 3.10+ with async/await throughout
  • FastAPI / Starlette ASGI for the HTTP layer
  • PostgreSQL for primary storage + pgvector for semantic search
  • Redis for caching and session store
  • Docker Compose for one-command deploy
  • E2B for sandboxed code execution
  • OAuth 1.0a / 2.0 / OIDC / WebAuthn / FIDO2 for auth

​
Who is this for?

Individual users

Anyone who wants a private, AI-powered home for their health data — without handing it over to a third party

Developers

Engineers building personal-data AI products: health, finance, legal, devops, or anywhere else tools/ + skills/ can take you

​
Real-World Examples

Showcase conversations below use synthetic health data and some Theta-private tools.

​
Private Health Data Management & Chat

Query the same condition across multiple health systems, even across countries:

Knee Pain Progress

Track knee pain progression across providers

Cardiovascular Disorders

Monitor cardiovascular health across health systems
Combine device data, medical records, and self-reported notes:

Cardiovascular Summary

Summarize cardiovascular conditions across sources

Diabetes History & Progress

Track diabetes from devices, records, and other sources

PCP Visit Preparation

Generate a comprehensive summary and questions for your primary care visit

​
Deep Research for Personal Health

Figure out what’s actually helping or hurting you:

Blood Glucose Analysis

Find factors that help or harm your blood glucose

Symptoms & Feelings Drivers

Discover the main drivers behind your symptoms and mood

Diabetes Treatment Options

Research personalized options (meds, devices, lifestyle) for diabetes management

​
Mirobody as a Data Engine for Developers

Wearable manufacturers

Add AI chat over your device data without building the backend

Research applications

Drop custom tools into tools/ to deploy subject-facing research apps

Consumer health apps

Use Mirobody as the data + AI backend behind your product

Enterprise

Contact us for custom integrations and enterprise support

​
Theta Wellness: Mirobody in Production

Theta Wellness is our flagship application built on Mirobody — a personal health intelligence app that demonstrates what Mirobody can do at production scale. It connects 300+ devices, Apple/Google Health, and EHR records, and supports voice, image, file, and text as inputs. We’re open-sourcing the engine because the same architecture that powers a medical-grade health agent can power your finance analyzer, legal assistant, or devops bot. Swap the files in tools/ and you’re in a new vertical.

​
Next Steps

1

Get started

Follow the Quick Start to deploy locally with ./deploy.sh.
2

Configure providers

Set OAuth credentials for the devices you want — see Configuration.
3

Explore the API

Read the API Reference for Pulse, Chat, and MCP endpoints.
4

Add custom providers or tools

Use Provider Integration for new devices, or just drop a Python file into tools/.

​
Contributing

We welcome contributions — new providers, new tools, framework improvements, anything.

GitHub Repository

Source code, issues, and PRs

Contributing Guide

How to submit changes

​
Community & Support

Discord

Community discussions (coming soon)

GitHub Issues

Bug reports and feature requests

Mirobody Support

Direct technical support
This docs site covers Mirobody’s health vertical. For the open-source engine repo, see github.com/thetahealth/mirobody.