Know who asked
OCI IAM validates the signed-in identity before the agent workspace opens, linking each request to an authenticated subject.
Identity verified
Negative controls · sealed evidence · powered by OCI
Negative controls,
Negative controls,
fully evidenced.
Meridian runs six repeatable HR security scenarios—including semantic transformation, unverified mutation, embedded data-field instructions, and cross-request effects—then seals the identity, OCI Responses calls, SQL, Oracle results, and audit trail for review.
OCI IAM Verified identity
OCI Responses Observable function calls
Oracle Database Correlated execution evidence
From identity to evidence
Every action has a trace.
Meridian retains each model-generated SQL attempt—including syntax failures—and correlates it with the authenticated user, request, Oracle result, and independent evidence sources.
See what ran
OCI Responses selects function calls and generates SQL. Meridian records the exact statements and Oracle results for review.
Actions recordedVerify the outcome
A hash-linked journal, Oracle audit records, OCI Logging, exact results, and integrity baselines form a correlated evidence chain.
Evidence sealedTransparent demonstration architecture
Clear boundaries. Verifiable results.
This separation keeps the evidence easy to evaluate: SQL originates in the managed model and function workflow, while Oracle remains a conventional execution target. The demonstration uses synthetic HR data and clearly identifies components that are outside the current evidence path.
Correlated evidence available