How to use Claude to build SaaS with three coding roles
The author reports a deployed Telegram Ads SaaS with 24 interactive HTML pages, 60+ API endpoints, 311 migrations, Kubernetes infrastructure and Stripe and PayPal billing.
news.ycombinator.com
Prompt to try
Editorial adaptation, not the source author’s exact prompt. Replace bracketed inputs and review tool permissions first.
For [SaaS project], use separate [frontend], [backend] and [coordination] roles with explicit file boundaries. First build [page] with demo data for my review. Derive backend requirements from the approved UI, write [endpoint task document], resolve questions through [message directory], then connect real API calls.
Reach people exploring AI agents
See placements →What happened
The author reports a deployed Telegram Ads SaaS with 24 interactive HTML pages, 60+ API endpoints, 311 migrations, Kubernetes infrastructure and Stripe and PayPal billing.
Reported by yego on Hacker NewsHow the agent did it
- Directed the frontend agent to build pure HTML, CSS and JavaScript pages with demo data.
- Iterated on the interface until satisfied with its appearance.
- Extracted backend data requirements from the completed pages.
Where a person stepped in
The author directed the agents, reviewed interface iterations and kept the pipeline strictly sequential.
A person approved each step before the agent continued.
A solo SaaS builder described a three-agent Claude Code setup developed after months of iteration. An earlier attempt to delegate broadly to a CEO agent had produced an expanding organisation and excessive coordination instead of work. The revised setup constrained three Docker-contained agents with separate CLAUDE.md files and strict access boundaries.
The frontend agent owned HTML templates, JavaScript and CSS. The backend agent owned the Go API, database and migrations. The CEO handled strategy, content, marketing and task coordination, with access to documentation and a prohibition on writing code.
Work proceeded sequentially. The author directed the frontend agent to build a page using demo data and iterated until satisfied with its appearance. They then extracted the backend's data requirements from that finished page and placed an endpoint specification in docs/tasks/backend.md. Agents exchanged clarification messages through docs/messages/. Finally, the frontend replaced demo data with real API calls.
The author reported deploying growity.ai, a Telegram Ads automation SaaS, with 24 interactive HTML pages, more than 60 API endpoints, 311 database migrations, billing and a Kubernetes deployment. The account attributes this to one person directing the agents.
Who it's for: solo product builders who want explicit role boundaries and UI-led backend requirements.
Prompt
Suggested prompt (written by Agents In Use, not the author's)
Source: Hacker News · yego
Frequently asked questions
Did the agents work in parallel?
No. The author described a strictly sequential pipeline: frontend demo, requirements extraction, backend task, clarification and frontend API integration.
What stopped the CEO agent from adding complexity?
The revised setup constrained the CEO to strategy, content, marketing and coordination. Its key rule was that it could read code but never write it.
More details
Remaining steps
- Wrote endpoint tasks in docs/tasks/backend.md and coordinated clarifications through docs/messages/.
- Connected frontend pages to real API calls.
About this account
Three Claude Code agents handled frontend, backend and coordination within separate roles and workspace boundaries.
Platform: Hacker News. Attribution: https://news.ycombinator.com/item?id=47245374. Prompt: Suggested (written by Agents In Use, not the author's). Sourced by the Agents In Use pipeline.
Credit: yego on Hacker News
Source published 4 Mar 2026
Source checked 14 Oct 2026
Tools and topics
Claude Code · Docker · Markdown · Go · Kubernetes · PostgreSQL
SaaS · Agent roles · UI development
Editorial completeness: 100/100. Measures documented fields, not agent performance or verified outcomes.