Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.medera.info/llms.txt

Use this file to discover all available pages before exploring further.

Medera is the AI infrastructure platform for behavioral and mental health. It combines a canvas-based agent builder, multimodal clinical sensing (vocal acoustics + facial physiology + RDoC constructs), purpose-built medical reasoning, and a HIPAA-grade runtime — delivered through a single REST + WebSocket API and a typed JavaScript / Python SDK.
Medera’s goal is to be the most accurate, defensible, and clinician-ready AI platform for behavioral and mental health teams — without compromises on integration speed, evidence, or regulatory compliance.

​
What’s in the platform

Agent Builder Canvas

A visual workflow editor with 16 core node types + 8 BH-OS rails, hash-based draft dedup, semver versioning, immutable published snapshots, and sandbox-first deployment to sandbox.medera.info/{org}/{agent}-v{ver}.

Multimodal Sensing

Facial Physiological Engine (rPPG + HRV with SDNN, RMSSD, LF/HF, SD1, SD2), Vocal Acoustic Engine (Praat / Parselmouth — F0, jitter, shimmer, HNR, MFCC, clinical markers), and the RDoC Construct Computer (15 constructs across 5 domains).

Care Orchestrator

Multi-agent orchestration over 4 voice agents (Grace, Alex, Jordan, Taylor) and 6 workflow agents (PRIOR_AUTH, REFERRAL, PRESCRIPTION, LAB, CARE_PLAN, ESCALATION) with an event bus and a 30+ state workflow state machine.

HIPAA-grade runtime

RLS on every tenant-scoped table, AES-256-GCM PHI fields, Merkle-tree audit, purpose-based access (TREATMENT, PAYMENT, OPERATIONS, EMERGENCY), and immutable WORM audit on every deployment action.

​
Service topology

┌──────────────────────────────────────────────────────────────────────────┐
│                         FRONTEND (Port 5173)                             │
│   React 19 + Vite 7 + Tailwind 3 + Socket.IO + TensorFlow.js             │
└────────────────────────────────┬─────────────────────────────────────────┘
                                 │
                 ┌───────────────┴───────────────┐
                 ▼                               ▼
┌──────────────────────────────┐  ┌────────────────────────────────────────┐
│    BACKEND (Port 3001)       │  │      AI SERVICES (Port 8000)           │
│  Node.js/Express + Socket.IO │  │    Python/FastAPI + OpenAI + PyTorch   │
│  58 routes · 57 services     │  │  22 routers · multimodal engines · PA  │
│  18 middleware · RLS · BAA   │  │  Care Orchestrator · Clinical RAG      │
│  Workflow Engine + Bull queue│  │  Vapi voice orchestration              │
└──────────────┬───────────────┘  └──────────────┬─────────────────────────┘
               │                                  │
               ▼                                  ▼
┌──────────────────────────────────────────────────────────────────────────┐
│              POSTGRESQL 16 (5433) + REDIS (6380)                          │
└──────────────────────────────────────────────────────────────────────────┘

​
Common workflows

Build a canvas workflow

Drag a graph of agents, tools, and logic. Publish a semver version. Deploy to sandbox. Promote to production.

Run an ambient scribe

Open the clinical session WebSocket, stream encounter audio, get diarized transcript + SOAP draft + ICD-10 / CPT suggestions.

Deploy Grace for intake

Deploy the voice-first intake agent with a documented 7-state state machine and 4 function calls.

Generate a PA letter

Drive a 17-state PA workflow with payer-specific rules across EPA_API, CLEARINGHOUSE, PORTAL, FAX, or MANUAL submission.

​
What’s next

Console

Sign up, provision your tenant, generate credentials.

Authentication

OAuth 2.0 client credentials + Clerk JWT + API keys.

API Reference

Every endpoint across backend and AI services.
Questions about implementing Medera in your environment? Contact us.