Pricing
burntop is free and open source, distributed as a GitHub template with no stated paid tier, subscription, or hosting fee.
Learn moreburntop
A FastAPI, React and Postgres starter that fails the build when an AI agent cuts corners.
Intermediate API
What is burntop?
burntop is a developer tool. Open source FastAPI, React 19, and Postgres template with built-in auth, a reference CRUD module, and 200+ end-to-end tests that block agent-generated code with mocks or missing coverage from passing the build.
burntop is an open source full-stack template built for teams that generate most of their code with AI coding agents. It pairs a FastAPI and Postgres backend with a React 19 frontend, and ships with working authentication, a reference CRUD module, and more than 200 end-to-end tests that run against a real database and a live server rather than mocks. The core idea is enforcement: a route without test coverage, a mocked dependency, or an untested page fails the automated build, so an agent cannot quietly mark a feature done when it isn't. The template is aimed at developers who already lean on tools like Claude Code, Cursor, or Codex to scaffold applications and want guardrails around that workflow. Nested AGENTS.md files and a standards skill tell the agent how the codebase is organized and when a feature actually qualifies as finished, while the existing items module gives it a working pattern to copy for every new feature instead of inventing its own conventions. In practice, a developer creates a new repository from the GitHub template, starts the stack with Docker Compose, and either follows the quickstart manually or hands the README and AGENTS.md to an agent and lets it extend the app. Session auth with Google OAuth, password reset, and email verification are already implemented, along with a typed FastAPI backend that produces OpenAPI docs and a consistent error shape for the frontend to consume. Among boilerplate and starter-kit alternatives, burntop's distinguishing feature is the testing policy itself: banning mocking libraries and gating the build on real end-to-end coverage, rather than offering a generic scaffold and leaving verification up to whoever (or whatever) writes the next feature.
How do you use burntop?
1Install Docker and the GitHub CLI
These are the only two prerequisites needed before creating a project from the template.2Create a project from the template
Run the gh repo create command with the agusmdev/fullstack-ai-template flag, or use the 'Use this template' button on GitHub to get a fresh repository.3Copy the agent prompt and hand it to your agent
Use the provided prompt (create a new project from the template as described in its README, then open AGENTS.md and follow it) with Claude Code, Cursor, or Codex.
4Start the stack with Docker Compose
Copy the example environment file and run docker compose up to start Postgres, the API on port 9095, and the app on port 3000.5Build a feature by copying the items module
Point the agent at the existing items module as the pattern to follow for models, repository, routers, and tests when adding a new feature.6Run make ci before considering work done
The build only passes when static checks, backend end-to-end tests, and Playwright tests all succeed, which catches mocks or untested routes.
Pros and cons
Pros
- Enforces real end-to-end tests against Postgres and a live server instead of allowing mocksAI
- Comes with working session auth, Google OAuth, password reset, and email verification already builtAI
- Includes a reference CRUD module (items) that new features can copy for consistent patternsAI
- Pinned, modern stack (FastAPI, React 19, Postgres 18, Tailwind 4, shadcn/ui) with Docker Compose setupAI
- Free and open source, usable directly as a GitHub templateAI
Cons
- Requires comfort with Docker and the GitHub CLI, so it is not a no-code or beginner-friendly starting pointAI
- Does not send emails itself; a provider has to be wired in separately for reset and verification flowsAI
- No AI or LLM functionality exists inside the generated app, only in the agent workflow used to build itAI
- Strict no-mocks, full-coverage policy can slow down quick prototyping or exploratory changesAI
- No hosted service or support plans; it is a self-managed template without a vendor pricing or SLA structureAI
How much does burntop cost?
Support
Support runs through the project's GitHub issues page and an FAQ section on the site; general contact is listed as contact@techwarely.com.
Integrations
Includes Google OAuth for sign-in, Sentry for error tracking, GitHub Actions for CI, and Docker-based deployment; built to work with AI coding agents that read AGENTS.md, including Claude Code, Cursor, and Codex.
Features
Working session auth with Google OAuth, password reset and verification; a reference CRUD module with search, filters, and pagination; 200+ real end-to-end tests banning mocks; a typed FastAPI backend with OpenAPI docs; Docker Compose setup with roughly 60 make targets for CI.
Frequently asked questions about burntop
How much does burntop cost?
burntop is free and open source, distributed as a GitHub template with no stated paid tier, subscription, or hosting fee.Does burntop have an API?
Yes, burntop offers an API.How do you use burntop?
The walkthrough on this page covers 6 steps: 1. Install Docker and the GitHub CLI 2. Create a project from the template 3. Copy the agent prompt and hand it to your agent 4. Start the stack with Docker Compose 5. Build a feature by copying the items module 6. Run make ci before considering work done.What platforms does burntop support?
burntop is available on Web App.What does burntop integrate with?
Includes Google OAuth for sign-in, Sentry for error tracking, GitHub Actions for CI, and Docker-based deployment; built to work with AI coding agents that read AGENTS.md, including Claude Code, Cursor, and Codex.What are the limitations of burntop?
Requires comfort with Docker and the GitHub CLI, so it is not a no-code or beginner-friendly starting point. Does not send emails itself; a provider has to be wired in separately for reset and verification flows. No AI or LLM functionality exists inside the generated app, only in the agent workflow used to build it.
