What a coding agent produces in a multi-tenant SaaS codebase
Once the tenant model and billing design exist, much of a SaaS codebase is repetition around them, and that is the work we give to agents. An engineer writes a scoped task with acceptance criteria and the boundaries the agent must not cross, for example: add a CSV export to the reports screen, scoped to the current tenant, without touching the auth middleware or the billing module. The agent reads the existing schema and route handlers, then writes the CRUD screens, the admin views, the Drizzle migrations, the request validation on Hono endpoints and the tests around them. It runs the type checker, linter and test suite before proposing a change, and a senior engineer reviews every diff before it merges. Agents never merge their own work, and continuous integration re-runs the checks on every change. Reveal’s published case documents a connected Next.js web application, browser extension and admin dashboard using Hono, Drizzle ORM and PostgreSQL. It establishes product scope and stack, not which historical changes were authored by coding agents.





