Try the decision engine

Send an action, see the verdict, approve or reject it, then edit a record and watch the chain check catch it.

1. Send an agent action

This is a sandbox. It runs the same policy evaluator and redaction code as the production API. Evidence records are hashed field by field with SHA-256 (hash version 2), against a fixed set of five policies. Nothing is stored and no account is needed. The default for unmatched actions is BLOCK.

2. Decision

Pick an example and choose Evaluate action.

4. Evidence chain

Each decision becomes a record. Each record contains the hash of the one before it.