Getting Started
Build health agents on standardized data
Store source data, standardize health readings, and put AI to work on them.
Mirobody API turns messy health data into standardized, AI-ready records — then gives you two OpenAI-compatible surfaces to build agents on them. It’s the managed layer over the open-source Mirobody engine: we host storage, models, and keys; you point an OpenAI SDK at the base URL with an mb_live_* key.
The loop: upload → standardize → use with AI
Section titled “The loop: upload → standardize → use with AI”Match the endpoint to your data shape
Structured readings go to POST /v1/data. Lab PDFs, photos, and spreadsheets go to POST /v1/files, which stores the original, extracts its text, and standardizes any readings it finds. POST /v1/standardize runs that same standardization synchronously — for a dry-run preview, or on narrative text.
Every structured reading is standardized
Names resolve to LOINC deterministically (no LLM code-guessing), values normalize to UCUM units, every reading gets a FHIR mirror. "血糖(空腹)", "FBG" and "Glucose, fasting" become one series. See Standardization.
Put AI on top
Ask the Answers API for a grounded, evidence-cited answer — or build a full agent on the Agent API: your own tools, stored conversations, and the openai-agents SDK working out of the box.
Four shapes of data, one door each
Section titled “Four shapes of data, one door each”Health data arrives in four shapes — each has exactly one door in:
| Data shape | Typical source | How it goes in | Endpoint |
|---|---|---|---|
| Structured readings | Devices / wearables (you own the integration — write daily aggregates), manual entries | Structured records | POST /v1/data — pre-aggregate high-frequency samples; episodes (sleep, workouts) carry time + end_time. See the device-data cookbook. |
| Files / photos | Lab reports, checkup PDFs, phone photos | File upload | POST /v1/files — stores the original, extracts its text, and pulls out standardized readings automatically. POST /v1/standardize runs the same extraction synchronously — for example as a dry-run preview. |
| Narrative with readings | “headache all day, temperature was 38.2 °C”, dictated notes | Narrative text | POST /v1/standardize — the quantifiable readings are extracted; the narrative around them is dropped. |
| Purely subjective journal | “dizzy and a headache all afternoon”, mood notes | Single-turn agent call | POST /v1/responses with store: true — readings and durable memories are extracted from the entry. See the Journaling recipe. |
Mirobody stores source data, standardizes structured readings, and makes both available to AI. You decide how data is collected and prepared before it reaches the API.
Two API surfaces, one engine
Section titled “Two API surfaces, one engine”POST /v1/responses (OpenAI Responses-compatible). Your function tools, previous_response_id / session_id state, response.* streaming. The openai-agents SDK needs only a new base URL.
POST /v1/chat/completions. A closed, grounded completion — one question, one evidence-backed answer. Drop-in for any OpenAI SDK, and ideal as a tool inside your own agent.
Not sure? Choose your API.
Why Mirobody
Section titled “Why Mirobody”- Grounded, not plausible — answers can include
health_records,citations, and the server tool steps used to gather evidence. - Standardized structured data — LOINC + UCUM + FHIR on records written through
/v1/dataor stored by/v1/standardize; your analytics and the agent read the same series. - OpenAI-compatible twice over — Chat Completions and Responses protocols; existing SDKs and agent frameworks just work.
- Multi-tenant by design — one key, an isolated Subject per end-user; per-Subject right-to-be-forgotten deletes.
- Open at the core — the hosted API runs the same open-source engine.
Two ways to run Mirobody
Section titled “Two ways to run Mirobody”- Hosted API (this platform) — an
mb_live_*key against our managed clusters. Start with the Quickstart. - Open source —
git clonethe Mirobody engine, build it with./build.sh, and run it yourself.
Start here
Section titled “Start here”Key → data → standardized records → your first agent, in minutes.
Answers vs Agent — which surface fits.
Auth, multi-tenancy, retention, errors, every endpoint.
Which cluster to use and what’s live today.