Demo build. Every business shown is fictional, data is synthetic and partner integrations are simulated.What is real here

Missions

The Agent Lab tests your business with declared agents running missions written in a small, versioned DSL.

A mission is a task a customer's agent might perform, written as data: an intent in the customer's own words, a persona (locale and device), constraints, the conversion it should reach and the expectations that decide pass or fail. A nightly lab plan expands each mission into runs across surfaces and agents, within the budget you set, and reports outcome rates with confidence intervals and a trend gate.

  • Generic missions are generated from your Business Graph (categories, offerings, locations, policies), so the lab runs for a business with no industry pack at all. Packs add their own missions.
  • Release gate: a targeted subset runs against a staged projection before it publishes; a regression blocks the publish with links to the transcripts.
  • Validation: agora validate mission <file> checks a mission against the schema before you commit it.

Lab agents always declare themselves and follow the rules on the lab transparency page.

id: find-and-compare
intent: "Find {offering_category} near {location} available {timing}, and compare two options"
persona: { locale: "{locale}", device: mobile }
constraints: ["max_price: {budget}"]
conversion: inquiry
expect:
  - reaches: conversion_started
  - quotes_policy: cancellation

Source: docs/portal/docs/missions.md