From FHIR findings to verified improvements.
Inspect the resource and requirement behind a finding, agree who will investigate it, and verify the result after a source correction. Keep the remaining findings ready for a release or data handover.
Check, understand, improve
Understand the problem before you change the data.
Find resources in Browse, follow their references and read profile requirements beside the data. When a check finds a problem, trace it to the affected field before changing the source.
- Check conformance
- Check resources against the selected profiles, terminology and project rules.
- Understand data quality
- Read completeness, uniqueness and reference checks with their evaluated population.
- Verify the change
- Give the finding an owner, repeat compatible checks after correction and preserve the result.
New, resolved or changed. Follow every check.
The Validation pane keeps finding changes beside their field and requirement. Watch one synthetic Observation move through a correction, an intentional regression and recovery.
5:39 · English captions · Recorded 22 October 2026 · Synthetic data
A reviewed Claude Code change is written to Aidbox and read back before Records checks the source again. The regression is intentional. A separate unresolved extension finding remains visible.
Edited for time; CLI transcript is labeled in the film. Narration uses a synthetic voice. Technical checks do not establish clinical correctness.
Public demo
Try a temporary correction
Sandbox compares with its first complete check. Resource revalidation compares the last two complete checks; reloading clears this temporary comparison.
Records
Compare recorded runs
Select a baseline run and review recorded issue types and occurrence changes. The stored report keeps the comparison available for handover.
Changed or incomplete validation inputs can prevent a comparison. Observed finding changes do not by themselves prove that a resource edit caused every difference.
Find the data-quality issue behind the number.
In this synthetic dataset, two Patients share an identifier. Records shows the affected resources and the evaluated population, so the team can investigate the duplicate and check the result after correction.
Choose completeness and reference checks alongside uniqueness to answer the receiving team’s quality question.
How Records fits
Records measures beside your FHIR system.
Your deployment defines the permitted scope. Records reads alongside your FHIR infrastructure and retains derived findings, comparisons and evidence for the receiving team.
FHIR source
Permitted server, environment, and resource scope.
Validation basis
FHIR release, packages, terminology, rules, and thresholds.
Records core
Validate, compare, investigate, and preserve.
Decision signal
Threshold, delta, and concrete findings.
Evidence
Run basis, timestamps, integrity, and export.
Start with the question your team needs answered.
Evaluate a delivery, inspect the workflow, or integrate validation into the tools you already use.
Evaluate your workflow
Plan a scoped evaluation
Bring one delivery, its profile basis, and a quality question. Review findings and repeat the checks after a correction.
Inspect the product
Explore Records workflows
Follow a finding, compare profile requirements, or investigate the quality of a dataset.
Choose the decision
Find the use case
Start from the release, comparison, reporting, research, or automation decision you need to support.
Integrate validation
Choose a developer surface
Run the CLI, embed the engine, use Records Agent Tools in Claude Code or Codex, or expose Records through MCP.
Quality & evidence
Explore research data quality
Trace a dataset indicator to its affected resources and verify the result after a source correction.
Quality & evidence
Explore evidence handover
Review the checked scope, findings and validation basis in the report sent to the receiving team.