Open SaaS
wasp-lang’s MIT-licensed SaaS boilerplate: “A 100% free modern JS SaaS boilerplate (React, NodeJS, Prisma). Full-featured: Auth (email, google, github, slack, MS), Email sending, Background jobs, Landing page, Payments (Stripe, Polar.sh), Shadcn UI, S3 file upload.”
15.3k stars, 1.8k forks, 101 open issues as re-verified on 2026-08-12. (The hub’s park record from
2026-08-10 said 15,217 stars / 1,818 forks / 114 open issues; the star and fork counts round to the
same figures and the issue count has moved. Recorded per ../HUB.md — the park record is a lead, the
primary is the finding.)
T1 as the project’s own repository, with the obvious caveat: a project describing itself. What it is good for is what is offered, and every claim below is the README’s.
What is pre-decided
This is the point of the artifact — the decisions are made before you start. Auth against email, Google, GitHub, Slack and Microsoft. Payments through Stripe, Polar.sh or Lemon Squeezy. Background jobs and cron via Wasp Jobs. A ShadCN admin dashboard. AWS S3 uploads. Email through SendGrid, Mailgun or SMTP. Analytics through Plausible or Google Analytics. End-to-end tests in Playwright. Docs as an Astro Starlight site. Deployment to Railway or Fly.io, or by hand.
Note what that list is: a dozen commitments a team would otherwise argue about for a week each, taken as a bundle. This spoke’s question is who decides an application’s structure, and here the answer is a template author whose name the adopting team may never learn.
The framework underneath is the more interesting half
Open SaaS is a template for wasp, and the architectural claim lives there rather than here.
Wasp is a declarative full-stack framework with a compiler: you write main.wasp.ts describing the
app at a high level and “Wasp compiler generates the whole source of your web app in the target stack:
front-end, back-end and deployment.”
AI-ready, and why that is not a route
The README advertises being “AI-Ready with tailored AGENTS.md, skills, and Claude Code plugin” and
“ready to work with your favorite AI coding tool or agent (Claude Code, Cursor, Codex, OpenCode,
etc.).” That surface is a bolt-on to a web template. Per this spoke’s CLAUDE.md boundary, it does not
move the subject toward ../agentic-tooling-wiki — but it is worth one observation the corpus can use:
a boilerplate whose selling point is that an agent will find its conventions legible is a claim about
architecture-as-interface, aimed at a reader who is not human.
Why it counted toward the spin-out, and the objection
The hub’s own park record argued both sides and refused to decide. For: it answers the founding question from the third possible direction — the framework deciding, which the cluster was missing. Against: genre. The other two founding sources are architecture guides that argue a thesis; this embodies one silently, and the same cluster had already excluded a build-along tutorial on exactly that distinction.
The objection is recorded rather than dissolved. The curator approved the spin-out on 2026-08-12 with both readings in front of them. What settles the genre question in practice is that the template’s choices are inspectable: you can read what it decided, which is what an architecture source is for.
Related
wasp · application-architecture · boilerplate · micro-frontends-migration · flutter-large-app-structure · synthesis