Integration architecture

Tebra Integration: How MedReception AI Connects to Tebra Clinical and Practice Management

MedReception AI has run a live production Tebra integration since May 25, 2026. This page documents the architecture, the read and write surface, the tenant onboarding flow, and the reliability posture for practices considering connecting their Tebra instance.

Live in production, 99.7% write success

Addressable market

50K

Tebra practices across the US, independent and small group dominant

Live since

May 2026

Production cutover on 2026 05 25 with the first paying tenant

Round trip latency

4.2s

Average AI to Tebra to AI write confirmation

Write success rate

99.7%

Appointment writes confirmed by Tebra within retry budget

Tebra product overview

What Tebra Is and Why It Matters to Independent Practices

Tebra is the merged entity created when Kareo, a long standing ambulatory EHR and billing platform, combined with PatientPop, a growth and reputation suite. The combined product line covers clinical charting, practice management, billing, patient acquisition, and reputation in a single tenant. It is one of the most common EHRs in independent primary care, behavioral health, podiatry, OB GYN, and specialty solo and small group practices.

Clinical and Charting

Note templates with structured and free text sections
eRx with controlled substance support
Encounter status tracking from scheduled through billed
Labs and imaging order management
Problem list, allergies, medication reconciliation

Practice Management

Multi provider scheduling with visit type rules
Patient demographics and household relationships
Insurance capture with primary and secondary plans
Copay collection and patient balance tracking
Eligibility checks and authorization workflow

Acquisition (PatientPop heritage)

Practice website and SEO tooling
Online reputation and review aggregation
Patient intake forms with EHR sync
Marketing automation and recall campaigns
Lead capture from web forms into the practice

Billing and RCM

Charge capture and superbill generation
Clearinghouse claim submission
ERA posting and denial workflow
Patient statements and payment plans
Reporting on AR aging and collection ratios

Integration architecture

REST API Plus a Browser Side Session Keeper

Tebra exposes a documented REST surface for the majority of read and write operations. A subset of high value workflows, particularly some of the legacy Kareo scheduling screens and certain billing actions, still require a logged in browser context. MedReception AI runs a hybrid architecture that uses the REST API first and falls back to a dedicated EC2 hosted session keeper for screens that the API does not yet expose.

Layer 1: Tebra REST API

OAuth 2.0 tenant authentication with refresh tokens
Practice scoped credentials with least privilege scopes
Pagination aware reads for patients, providers, appointments
Idempotent writes for appointment create and patient demographics
Rate limit aware client with exponential backoff and jitter

Layer 2: Session Keeper (EC2)

Long lived Playwright session on 3.82.129.176
Holds an authenticated browser context per tenant
Routes browser only workflows over an HTTPS RPC channel
Auto reauth on session expiry with 1 attempt cap and alert
Per tenant isolation, no shared cookies across practices

Routing Logic

Action dispatcher probes the REST endpoint first
Falls back to the keeper only when the REST path is unavailable
Same calling contract for the upstream AI receptionist
Latency budget enforced end to end at 8 seconds
Telemetry tags every call with the path that handled it

HIPAA and Tenant Isolation

BAA in place with the practice before any production write
Per tenant secrets in AWS Secrets Manager, scoped IAM
Audit log of every read and write with correlation id
PHI minimization, only fields needed for the workflow are read
Encrypted in transit and at rest, key rotation on schedule

Read surface

What MedReception AI Reads from Tebra

The AI receptionist reads only the fields it needs to answer a caller, route them correctly, or build a booking. Each read is logged with the calling intent so the practice can audit what the AI saw and when.

Schedule and Availability

Provider availability windows and block reasons
Existing appointments within the booking horizon
Visit type duration rules and template constraints
Multi location filtering by clinic site
Holiday and PTO blocks

Patient Records

Patient lookup by name, DOB, phone, MRN
Demographics for callback and SMS confirmations
Communication preferences and consent flags
Active and inactive status
Linked household members for family booking

Insurance

Primary and secondary plan summary
Member ID, group, and effective dates
Last eligibility check result if cached
Copay and deductible snapshot when available
Plan acceptance flag for the booked provider

Clinical Context

eRx history for refill triage
Encounter and note status, scheduled through signed
Open orders for labs and imaging
Allergies and medication list summary
Last visit date and provider for continuity

Write surface

What MedReception AI Writes to Tebra

Writes are explicit, narrow, and reversible. Every write returns a Tebra confirmation id that is mirrored into the MedReception AI audit log. The practice retains full control through Tebra audit and can reverse any AI action through normal Tebra workflows.

Appointments

Create appointment with provider, visit type, location
Reschedule and cancel with reason code
Attach booking notes from the call transcript summary
Respect provider template rules and overbook policy
Idempotent on the Tebra appointment id to prevent duplicates

Patient Demographics

New patient creation with required fields validated up front
Update phone, email, preferred contact method
Update address and emergency contact
Diff aware updates, no overwrite of fields the AI did not touch
Soft duplicate detection before create

Insurance and Copay

Capture new insurance plan with payer, member id, group
Update primary versus secondary ordering
Record copay collected over phone with Stripe receipt
Attach proof of card on file when the practice opts in
Flag eligibility recheck if plan changed

eRx and Tasks

Create refill triage tasks routed to the prescribing provider
Attach call transcript summary to the task
Set urgency based on medication class and last fill date
Never auto sign or auto release prescriptions
Hand off to clinician for any controlled substance request

Tenant onboarding flow

From Connection Wizard to Production Cutover

A new Tebra practice is fully connected and validated in a single working session. The flow runs through a guided wizard that handles credentials, OAuth, sandbox validation, and production cutover.

Step 1: Tebra Connection Wizard

The MedReception AI portal walks the practice through Tebra API enablement and shows exactly which scopes to grant. The wizard validates each scope as it is granted.

Step 2: OAuth Handshake

The practice signs in through Tebra, consents to the requested scopes, and the refresh token is stored per tenant in AWS Secrets Manager. No credentials live in the application database.

Step 3: Sandbox Test Suite

A scripted suite runs against a Tebra sandbox or a designated test patient. It exercises every read and write the AI will perform, including create, reschedule, cancel, insurance capture, and refill task. The practice reviews the results before signing off.

Step 4: Provider and Visit Type Mapping

The wizard pulls the live provider roster and visit type catalog from Tebra. The practice confirms which providers the AI can book, which visit types map to which scheduling rules, and which locations are in scope.

Step 5: Production Cutover

The practice flips a single switch from sandbox to production. The AI receptionist begins routing live calls into Tebra with full audit logging and the practice retains a one click rollback for the first 30 days.

Production reliability

Live Since 2026 05 25 with a Hardened Operations Posture

Session Keeper on EC2

The keeper runs on 3.82.129.176 with a watchdog that detects auth state, restarts the browser context on demand, and posts to a centralized alert manager on any failure.

One attempt cap on reauth, then alert

Write Confirmation Discipline

Every appointment write is re read from Tebra before the AI confirms the booking to the caller. The 99.7% success rate reflects confirmed writes, not optimistic acks.

Read after write verifies state

Centralized Alerting

A shared alert manager dedupes Tebra incidents, escalates from warn to critical, and posts broken to healthy transitions to the on call channel with the tenant id and the failing endpoint.

Broken to healthy notifications close the loop

For Tebra practices

See the Tebra AI Receptionist in Action

Book a working demo that calls into a live Tebra sandbox, books a real appointment, and writes the patient back into your test environment. You see the round trip end to end.

Book a Tebra demo

Pricing

Per Provider Pricing, No Tebra Surcharge

MedReception AI prices the receptionist by provider seat. The Tebra integration is included, not a separate line item. Sandbox and production both run on the same plan.

See pricing

Related resources