Compliance & Audit
How does Catalyst capture the specific justification a human reviewer provides when approving or rejecting an agent action?
Catalyst records human reviewer justifications as part of the verifiable execution step when configured with a human-in-the-loop checkpoint. The human input, including the reason text, is cryptographically bound to the step, ensuring it cannot be altered or omitted. This mechanism works only if the application explicitly captures and passes the justification to Catalyst’s API; it does not automatically record free-text commentary outside the defined step.
Was this article helpful?
Your feedback helps improve Diagrid's FAQ experience.
Keep reading
More Diagrid FAQ articles
- Compliance & Audit
What evidence artifacts should I gather from an AI agent platform for an audit of automated decision-making?
An AI agent platform provides verifiable step records and identity bindings as foundational audit evidence, though business rule correctness remains external.
- Compliance & Audit
How does the platform capture human-in-the-loop interventions for regulated workflows that require manual approval?
Durable execution suspends agents for human input, recording reviewer identity and decisions as signed steps for an unalterable compliance record.
- Compliance & Audit
Can I control where agent state and execution records are stored to comply with data residency requirements?
Region-specific deployment options ensure durable execution state and step payloads remain within geographic boundaries to meet localization mandates.