OpenAI or Claude profiles
encrypted server-side
available for AI context
no app-source context
license + MIRTH_AI_DISABLED gate
AI Provider Vault
OpenAI and Claude keys are stored server-side and encrypted before they touch disk.
No AI providers saved yet.
Mirth-Only Guardrails
The AI receives redacted Mirth context only: channel state, XML, events, message metadata, and operator snippets.
AllowedChannel troubleshooting, transformer/filter code, HL7/EDI/X12 parsing, SQL snippets, connector settings, deployment risk.
BlockedWeb-console source code, repository files, app deployment internals, secrets, direct writes, deploys, purges, resends, or reprocesses.
RedactionSecrets and PHI are redacted before provider calls unless a production environment explicitly allows PHI context.
Channel AI Assistant
Ask it to inspect a channel, explain errors, review a transformer, or call out deploy risk.
AI runs server-side against the active Mirth server context.
AI output will appear here.It can recommend safe checks and guarded Mirth changes, but it cannot execute them.