Skip to content

Agent Business Lifecycle

The lifecycle keeps product intent, executable behavior, commercial terms, and operational evidence connected without rewriting history.

State the business promise, intended consumer, and success boundary. A clear promise makes later schema, surface, and pricing decisions easier to review.

Implement the capability with a prompt, workflow, or connected engine. Define the input and output schemas, runtime limits, provider configuration, and failure behavior.

Choose Human SaaS, Agent SaaS, or Dual. Configure the surfaces, visibility, locale, authentication, origin policy, and pricing intent required by that mode.

Test representative successful, invalid, empty, timeout, and provider-failure cases. Confirm that public clients receive useful errors without secrets or private provider data.

Create a Capability Release that fixes the executable version and trust contract. Activation makes that release eligible for publication and runtime traffic.

Project the active release to one or more channels. Each Publication identifies its channel and the release and commercial references used there.

Attach an Offer and settlement policy appropriate to the buyer and channel. Keep price, quota, payment, refund, and allocation terms explicit rather than inferring them from the current UI.

Monitor verified invocations, latency, failures, revenue, cost, margin, refunds, and audit evidence. Use those signals to decide whether to keep the release, create a new one, revise an Offer, or change a publication.

Releases and commercial contracts are historical evidence. If live behavior changes in place, a past invocation can no longer be explained reliably. New behavior therefore creates a new release; changed buyer terms create a new Offer or settlement policy.

The Agent Business remains the stable identity that connects these versions over time.