Killing the Clipboard: Building an Autonomous Healthcare Agent in 5 Minutes
This Tuesday at HL7 FHIR DevDays in Minneapolis, I am doing a live demo of something I have been building privately for the last few months.
We talk a lot about healthcare interoperability, but the actual user experience for the average patient is still broken.
Yesterday afternoon, I had a routine primary care visit for blood work. This morning, the lab results hit my inbox.
Instead of logging into a siloed patient portal or downloading another proprietary app, I opened a Telegram chat with my personal health advisor agent, Sally.
The stack: OpenClaw running on Claude, layered with HealthClaw.io guardrails for automated PHI redaction and data quality validation.
While standing in my driveway setting up a neighborhood garage sale, I asked the agent to pull the labs.
The problem is that my provider uses Medent a niche EHR system that is completely isolated from major TEFCA networks or national CMS-aligned directories.
Within five minutes, the agent executed the entire sequence:
Located the Medent endpoint.
Authorized the secure data pull.
Ingested the unstructured lab values.
Compared the new lipid numbers against my historical trends.
Before I finished setting up the garage sale tables, the agent confirmed exactly how the lab results looked, verifying that my cholesterol medication is working perfectly.
This matches the exact engineering thesis behind Josh Mandel’s recent health skills QR link architecture proposal for the “Kill the Clipboard” initiative.
We do not need more administrative portals. We need autonomous, context-aware agents that have the exact tooling and guardrails required to handle legacy data silos on behalf of the consumer.
I am running a live, end-to-end demo of this exact autonomous agent architecture this Tuesday at HL7 FHIR DevDays in Minneapolis.
If you are a builder working on agentic AI, localized data privacy, or FHIR data liquidity, let’s connect.
#interop #AIagents #fhir #devdays

Discussion about this post