Model-assisted identity decisions

Steerpast separates deterministic identity, risk, and policy signals from model reasoning. DecisionAgent can use Claude through the Anthropic API, validate the response, and pass the result into remediation and audit.

1. Identity context is assembled.
2. Risk and policy constraints are evaluated.
3. Claude can synthesize an outcome, rationale, and confidence.
4. The response is validated before entering remediation.
5. The decision provider and evidence remain inspectable.

← Back to Steerpast · View demo