DocsStart
Use cases
Four kinds of app planned call by call: a fitness coach, a support agent, a document back office and an engineering agent, with the day's answers.
Four kinds of app, each as the calls a coding agent would make while planning it, in order, and what must hold of each answer. Each page shows what the day's catalogs answer, and the catalog versions it was made from.
- Fitness coachA consumer app on iOS and Android: a food photo turned into macros, a chat coach, a twelve-week programme and a weekly summary. It reads Apple Health, Health Connect, Google Fit, Strava, Garmin, Fitbit, WHOOP, Oura and MyFitnessPal, and is built with Supabase, RevenueCat and Expo.
- Support agentA hosted agent, run for many customers, that reads support tickets and writes replies. It works in Zendesk, Intercom, Slack, Notion, Shopify and Stripe, and searches the help documents.
- Document back officeNo chat. Receipts and invoices are read each night and written to QuickBooks or Xero, with bank data through Plaid and files from Google Drive and Gmail.
- Engineering agentA local agent in Claude Code that works across GitHub, Linear, Postgres, Sentry and Slack and runs code. The developer already holds E2B and Vercel accounts, and wants a model for the agent on Anthropic or OpenAI.
The same files are the product's acceptance cases: run on fixed catalogs by the tests, and on a real catalog before a deploy, where each rule is checked. What a case cannot test is whether an agent asks the right question; see Goals.