Technical Documentation

Architecture, features, supported stacks, AI providers, and the product roadmap. Everything you need to understand how SaaSClaw works under the hood.

What is SaaSClaw? #

SaaSClaw is a self-hosted AI app builder. You describe the application you want in plain English, and an AI wizard agent plans the architecture, writes every file, builds the project, and deploys it to a live URL — all in a single session.

No boilerplate. No manual configuration. No DevOps pipeline to set up. You give a prompt, the agent gives you a running application with its own domain, SSL certificate, and production-ready infrastructure.

The core idea is simple: one prompt to a live URL. Whether you need a landing page, a React dashboard, a Django web app with a database, or a FastAPI backend, SaaSClaw handles the entire lifecycle from concept to production deployment.

SaaSClaw is self-hosted and open source (AGPL-3.0). You run it on your own server, connect your own API keys, and own every line of generated code. There's no vendor lock-in, no per-seat pricing, and no proprietary runtime. Your apps live in standard Git repos on your infrastructure.

How It Works — The Wizard Flow #

When you start a new project in the SaaSClaw Studio, you enter a prompt describing what you want to build. From there, the wizard agent executes a five-stage pipeline, each stage feeding into the next. Here's how it works:

1

Plan

The agent analyzes your prompt, determines the best technology stack for your requirements, and outlines the project architecture. It decides on frameworks, directory structure, database schema, and key components before writing any code.

2

Build

The agent writes every file your project needs — models, views, routes, controllers, templates, stylesheets, configuration files, and dependency manifests. It scaffolds the entire project from scratch, including auth setup, database migrations, and environment configuration.

3

Review

Before deploying, the agent reviews its own code for correctness, security issues, missing dependencies, and architectural problems. It catches what it can before moving forward, reducing the need for fix-up cycles.

4

Debug

If the build step produces errors — compilation failures, missing imports, syntax issues — the agent reads the error output, diagnoses the root cause, and applies fixes. This loop continues until the project builds cleanly or the agent determines the issue requires user input.

5

Ship

The project is deployed to a preview URL at <slug>.preview.saasclaw.ai. Once you're satisfied, you promote it to production at <slug>.saasclaw.ai with one click. SSL is provisioned automatically via Let's Encrypt, and the app runs behind Nginx with its own systemd service.

The entire process happens in your browser. You can watch the agent work in real-time via Server-Sent Events (SSE). Every file write, every build command, every fix — it's all streamed live. You can intervene at any point by chatting with the agent to adjust direction.

Supported Project Types #

SaaSClaw supports a range of project types, from static sites to full-stack web applications. The wizard agent picks the appropriate type based on your prompt, or you can specify it explicitly.

Type Best For Deploy Method
Static HTML Landing pages, documentation sites Nginx serves files directly
Vite React Single-page apps, dashboards, admin panels npm build → Nginx serves static output
Vite Vue Single-page apps, dashboards, admin panels npm build → Nginx serves static output
Vite Svelte Single-page apps, dashboards, admin panels npm build → Nginx serves static output
Next.js (SSR) Full-stack web applications with server rendering next build → standalone Node.js server
Django App Full web applications with database and auth Gunicorn + Nginx reverse proxy
Flask / HTMX Lightweight Python web apps, server-rendered UIs Gunicorn + Nginx reverse proxy
FastAPI Async Python APIs, microservices Uvicorn via Gunicorn + Nginx
Hugo Static sites, blogs, content-heavy sites hugo build → Nginx serves static output
.NET / C# ASP.NET Core web applications dotnet publish → systemd service

Every deployed project — regardless of type — gets the same production treatment:

  • Preview URL at <slug>.preview.saasclaw.ai for testing before going live
  • Production URL at <slug>.saasclaw.ai with automatic SSL via Let's Encrypt
  • Environment variables for secrets, API keys, and configuration
  • Deploy history with full rollback to any previous deployment
  • Git-based workflow — every deploy is a commit, so your code is always versioned
  • Custom domains — point your own domain at any deployed project with automatic SSL

AI Providers & Models #

SaaSClaw connects to external AI providers via OpenAI-compatible APIs. You bring your own API keys — there's no intermediary markup or usage surcharge. Each project can use a different model depending on the task.

Z.ai
GLM-5.2 — Full reasoning
OpenAI
GPT-5.3 Codex — Latest coding model
Anthropic
Claude Sonnet — Balanced coding
Claude Fable — Lightweight and fast
BYO keys. SaaSClaw doesn't resell API access. You add your own provider API keys in your account settings and choose which model each project uses. This means you have full control over costs, rate limits, and model selection.

Problems SaaSClaw Solves #

"I have an idea but can't code"
Non-technical founders, designers, and product managers can now ship real applications. Describe what you want in natural language and get a working product — not a wireframe, not a mockup, but actual running code deployed to production.
"I need an MVP fast"
What used to take weeks of setup, scaffolding, and configuration now takes minutes. Internal tools, prototypes, and landing pages go from concept to production in a single session. The wizard handles the tedious parts.
"I'm tired of boilerplate"
The agent scaffolds the entire project — authentication, database models, routing, API endpoints, styling, environment config. You start from a working application, not an empty repository with a README.
"DevOps is not my job"
One-click deploys with SSL certificates, Nginx configuration, systemd services, and environment variable management. The wizard handles infrastructure so you can focus on what your application actually does.
"I want to iterate faster"
Describe changes conversationally and see them reflected in your live application. The feedback loop is a chat message, not a pull request. Need to add a feature? Change the color scheme? Fix a bug? Just say so.
"I don't want to be locked in"
Self-hosted, open source under AGPL-3.0. Your generated code lives in standard Git repositories on your own server. No proprietary runtime, no per-seat licensing, no platform dependency. Take your code anywhere.

Infrastructure #

SaaSClaw runs on a single Hetzner dedicated server. Here's the full architecture:

View Architecture Diagram →
SaaSClaw Infrastructure Client Browser (Studio UI) Server (ubuntu-16gb-fsn1-1) Nginx (reverse proxy) Gunicorn Django workers (×8, sync) :8010 Wizard SSE View event_stream() generator PiBridge (pi_bridge.py) JSON lines over stdin/stdout --mode rpc (persistent) Pi Coding Agent v0.80.2 Celery Worker Async deploy tasks Django ORM AgentSession, Profile, Tokens Deploy Pipeline Storage PostgreSQL Project Repos + Worktrees Bare Git Repos External ZAI API (GLM-5.2) GitHub Preview / Production systemd + nginx Let's Encrypt OpenAI-compat API tool calls Pi runs as saasclaw user · Gunicorn workers :8010 sync · SSE via POST fetch · Nginx handles SSL termination

The diagram above shows the data flow from browser to deployment. Here's a breakdown of the key components:

Server: Single Hetzner node (ubuntu-16gb-fsn1-1) with 16 GB RAM
Reverse proxy: Nginx handles SSL termination via Let's Encrypt certificates
Application: Gunicorn running Django with 8 sync worker threads on port 8010
Async tasks: Celery Beat + Worker handles deploys, session cleanup, and background jobs
Database: PostgreSQL for all application data (sessions, profiles, tokens, projects)
Cache: Redis for caching, Celery broker, and session state
Per-project isolation: Each deployed project gets its own systemd service and Nginx location block
Deploy pipeline: Git-based with bare repos, worktrees, and rollback to any previous commit
File ownership: Project workspaces owned by saasclaw user; application code by nmoore

Roadmap #

This is where SaaSClaw is headed. We organize the roadmap by status — things actively being built, things planned for the near term, and ideas we're exploring. No vaporware here.

In Progress ● Active

Planned ◈ Upcoming

Exploring ○ Research

⚠ STAGING ENVIRONMENT — Go to Production →