From 30-minute manual screen recordings to one command, 2 minutes, zero retakes. The Compliance Evidence Recorder automates MDI telehealth partner onboarding.
Before Ziv Health could go live with MDI (MD Integrations) as its telehealth partner, the clinical team needed video evidence proving the platform meets 7 compliance requirements. Privacy policy visible. Terms acknowledged. Side effects shown before purchase. Consultation fee as a separate line item. The full intake questionnaire walked through. Treatment plan visible. And a clear way for patients to message their doctor.
Recording this manually meant opening a screen recorder, navigating to each page, scrolling to the right sections, filling out the 9-exchange intake questionnaire field by field, and hoping nothing went wrong. One missed scroll, one wrong click, and you start over. Each attempt took 30+ minutes.
And this had to be done for every telehealth category — weight loss, men's health, women's HRT, dermatology, hair loss, longevity, and more. Each category has a different intake flow. Manual recording doesn't scale.
MDI requires 7 evidence points: privacy policy, terms acknowledgment, consultation awareness, consultation fee, intake questionnaire, treatment plan, and doctor communication. Each must be visible in a single video.
Per-section video recording with title slides and banner narration. API proxy to forward intake calls from the dev server to Railway. ffmpeg concat for the final MP4.
Read the backend ConversationalIntakeService to extract every field ID, label, and option. Mapped 3 UI types (OptionCards, CheckboxGroup, InputFieldGroup) to Playwright selectors.
Dev mode uses zivrx-api-production.up.railway.app (Railway fallback), not api.zivhealth.com. Proxy was only intercepting one domain. Extended to cover both — exchanges started working.
Preferences exchange sends options+checkboxes+inputFields simultaneously but UI only shows one (infinite loop). MDI consent checkbox uses wrong fieldId. Built API injection pattern to bypass both.
All 7 requirements, all 9 intake exchanges, all fields filled. 6.5 MB MP4 uploaded to MDI submission portal. Waiting for clinical team Green Light.
Building CER uncovered two bugs in the intake flow that would have affected real patients. Both were worked around in the agent and flagged for fix.
The preferences exchange returns options, checkboxes, and input fields in a single response. The UI renders them in priority order (options > checkboxes > inputFields), so only options show. Clicking an option submits it, but the backend returns the same structure — creating an infinite loop.
The backend expects fieldId 'mdi_consent_confirm' but the frontend's checkbox handler sends 'mdi_consent' (the group name). The consent acknowledgment never registers in the conversation state, so the exchange never advances to the 'understand questions' sub-step.
The Compliance Evidence Recorder was built for MDI onboarding, but the pattern applies anywhere you need automated, reproducible video evidence of a web application.
MDI, OpenLoop, CareValidate — any partner that requires video evidence before going live. Record once per partner, per category.
Some states require video proof that your platform meets telehealth regulations. CER walks through state-specific requirements automatically.
Insurance networks reviewing telehealth platforms need to see the patient experience. CER records the full journey from product page through intake.
Investors evaluating a telehealth platform want to see the product. CER produces a polished walkthrough video in 2 minutes — no demo prep needed.
After every deploy, run CER to verify the patient-facing flow still works end-to-end. If any selector fails, the video shows exactly where it broke.
Run CER monthly and archive the videos. If a regulator asks 'what did your site look like on date X?', you have timestamped video evidence.
The agent drives a real Chromium browser through the patient-facing site at zivhealth.com. It fills real forms, clicks real buttons, scrolls to real content. The clinical reviewer sees exactly what a patient sees — because it is what a patient sees.
Each section is recorded as a separate video segment with a title slide (Evidence 1/7, 2/7...) and a banner narrating what is being shown. The segments are concatenated into a single MP4. No audio needed — the banners make the video self-documenting.
CER is included in every Lighthouse plan. Your first video records in 2 minutes.
Apply for Lighthouse →