Burntop is a full-stack starter template designed for developers who build applications with AI coding agents. It pairs a modern, pinned technology stack with strict rules and real end-to-end tests, so an agent can scaffold features quickly without faking a passing build.
What it offers
The template ships a working application rather than a blank scaffold. It includes session authentication with database-backed tokens, Google OAuth, password reset and email verification, plus a complete reference feature module that demonstrates every layer of the stack. Backend code follows a router, service and repository pattern, while the frontend uses typed React Query hooks with optimistic updates and rollback.
- FastAPI, SQLAlchemy 2, Postgres 18, Pydantic 2 and uv on the backend
- React 19, TanStack Start, Tailwind 4 and shadcn/ui on the frontend
- Docker Compose for local development and GitHub Actions for continuous integration
- Sentry integration with personal data stripped from events
Built for AI agents
The core idea is that an agent should be able to read the rules, copy existing patterns and verify its own work. Nested AGENTS.md files and a standards skill explain how the codebase is structured and when a task is complete. New features copy the items module, so the agent follows established conventions instead of inventing new ones.
Tests that cannot be faked
Every test runs against a real Postgres instance and a live server over HTTP. The suite includes 161 backend tests using Testcontainers and 42 Playwright tests against the full stack. Mocking libraries are banned by ruff, and a route without a test, a page without a spec or a model without a migration will fail the build. This makes it difficult for an agent to produce a green check that does not reflect real behavior.
Who it is for
Burntop suits developers and small teams who want to move fast with AI coding assistants while keeping production-grade discipline. It is a strong fit for anyone starting a new SaaS or internal tool who wants authentication, CRUD, typed APIs and a real test suite already in place. The template works with Claude Code, Cursor, Codex or any agent that reads AGENTS.md, and it can be extended with any AI provider the developer chooses.
