Product

Action Checkout

Receiver-side action assurance for external AI-agent requests. Evaluate a high-responsibility action under your own policy, preserve a record you can replay, and verify the Target’s separately issued outcome.

Company

First Instinct

Product family

Action Checkout

Current private beta

Refund Request Shadow

Components

What the product is made of

These describe behaviour and boundaries. First Instinct does not publish its enabling implementation on this site.

Action Proposal

The external action to be evaluated, with its intent and idempotency identity.

Evidence inputs

Source-attributed facts already available about the actor, mandate, runtime and payment context.

Receiver Policy

Your own rules. First Instinct evaluates against them; it does not define them.

Acceptance Evaluation Record

A reproducible record of the decision and the reasons behind it.

Target Receipt

The outcome the Target system separately issues and signs for what it actually did.

Replay

Re-run a past evaluation against its preserved evidence to reproduce the judgement.

Never own what you adjudicate

First Instinct consumes evidence; it does not own the sides it judges

Upstream evidence

Identity, mandate, runtime, payment and carrier evidence — signed by others.

First Instinct

Receiver-side evaluation under your policy.

Target-owned outcome

The Target signs what it actually accepted or completed.

Boundaries

First Instinct does not:

  • issue identities;
  • operate a naming registry;
  • hold customer credentials;
  • define customer policy;
  • move money;
  • execute the target action;
  • sign the target outcome;