WEBVTT

NOTE Edited, sampled synthetic screen recording; scripted; no external model call.
NOTE Source fb2473246c5900ec8e714358dacb36a3167d4602; MCP 0.5.1.

1
00:00:00.000 --> 00:00:08.000
A fixed synthetic walkthrough. No external model is called.
The browser approval and FHIR persistence paths are real.

2
00:00:08.000 --> 00:00:22.000
The proposed 72 bpm Observation is waiting for a human decision.
A separate FHIR read still finds the same seven IDs.

3
00:00:22.000 --> 00:00:27.000
Review the FHIR preview: patient, heart-rate code, quantity, and units.
Persistence still requires approval.

4
00:00:27.000 --> 00:00:32.000
The preview binds this patient to 72 bpm, with LOINC 8867-4 and UCUM /min.
Nothing has been saved.

5
00:00:32.000 --> 00:00:36.000
Approve &amp; save sends the reviewed proposal to FHIR.
The pane keeps the last direct snapshot until the next query.

6
00:00:36.000 --> 00:00:48.000
A new direct FHIR read finds exactly one added Observation.
Its patient, quantity, LOINC code, UCUM unit, and AIAST label pass the checks.

7
00:00:48.000 --> 00:01:00.000
Submit again: scripted mode repeats the same fixed proposal.
While it waits, the direct FHIR query still returns the eight approved-state IDs.

8
00:01:00.000 --> 00:01:04.000
Cancel &amp; revise declines this proposal.
Now repeat the independent backend query to check the result.

9
00:01:04.000 --> 00:01:15.000
The final FHIR read returns the same eight IDs as after approval: zero added, zero removed.
Canceling the second proposal created no additional Observation.

10
00:01:15.000 --> 00:01:30.000
Try the pinned synthetic walkthrough: lastehr.com/docs/quickstart
This checks persistence mechanics, not clinical correctness, model quality, or production authorization.
