How do I prove what an AI agent did after the fact?

#The question

If you are asking "How do I prove what an AI agent did after the fact?", you want the guarantee enforced by the runtime, not by convention. Swirls does that: each execution is bound to its deployed definition and scoped runtime identity.

#Who's asking

Security / compliance owner. Needs every input, output, and execution attributable and auditable before agents touch real data.

#Why Swirls is a fit

The audit log is append-only, and every operational event carries the identity of the execution that produced it. The answer to "what did the agent do" is a record you export, not a story you reconstruct from scattered logs.

User and machine executions carry identity, and workflows can include durable human review checkpoints. Workflows are declarative, repeatable, and auditable by construction.

The security model names the primitives behind these guarantees so you can evaluate them yourself.