hourzero.dev: Ship your AI chatbot product in days, not months

Hi all, I'm Bogdan, and I built hourzero.dev mainly to ease my work as a software dev freelancer, but also because I kept noticing something in the AI job market. 90% of AI work on platforms like Upwork comes down to a chatbot. Not always called that — sometimes it's "an AI assistant trained on our docs", sometimes it's "a private ChatGPT for our employees", sometimes it's "an agent that can search our knowledge base". But strip it back and you're almost always building the same thing: a streaming chat interface with auth, some form of billing or access control, and a layer of tooling on top. The problem is that the open-source templates (Vercel's AI Chatbot, Morphic, etc.) are great for learning but not so much for shipping. No billing. No rate limiting. No token budgets. No file management. No org management. You spend 6 weeks building infrastructure before you've written a single line of the product itself. And here's the thing about 2026 — code is becoming a commodity. AI IDEs write it. AI models suggest it. As it was before, the value is not in writing boilerplate code. But the bar for what "is" boilerplate has changed and is much higher. hourzero.dev is the foundation I built for myself and I'm now sharing with you. I've even deployed and used this as a ChatGPT replacement for myself. I own my stack and, most importantly, my data. What you get out of the box: → Streaming chat with multi-model support (OpenAI, Anthropic, Gemini or any other model — switch with a config change) → Bespoke chat-tread compaction logic that plays nice with prompt caching's cost savings. You're not locked in to Anthropic's API, apply to any model. → Better-Auth — you own your auth, your users, your data. Not locked into any auth vendor. Comes with email auth, gmail OAuth, orgs, teams, role-based access & management, password reset / update, account deletion → S3-compatible file uploads using presigned URLs — zero bandwidth costs (SAFELY upload direct-to-bucket), works with Railway, AWS S3, Cloudflare R2, or anything S3-compatible. I do want to emphasize SAFETY. You don't have to worry whether you’ve accidentally built a public “upload anything” endpoint with extra steps :) → Share conversations and files among users / team members / org members → Stripe subscriptions with token limits and real time usage tracking per user / per team / per org → Upstash Redis rate limiting — nearly free at any scale → Resend for emails → Tool use: web search (SERP, snippet search, webpage retrieval), code execution, memory tool — all configurable → Persistent cross-conversation memory with a user-managed memory bank → Postgres + Drizzle ORM — one env variable change to switch between Supabase, Neon, PlanetScale, or your own instance → Vercel 1-click deploy or Docker for literally anywhere else → Fully typed TypeScript throughout, Next.js App Router, shadcn/ui, tRPC, tanstack + other goodies Who it's for: Freelancers who build AI chatbot products for clients and want to ship in days instead of weeks. Indie developers building their first AI SaaS. Teams that need a multitenant chatbot foundation with org management and RBAC already built-in. Anyone who's been asked to build "something like ChatGPT but just for our company" and doesn't want to start from scratch for the fifth time. One-time purchase. Lifetime updates. Unlimited projects. I'd genuinely love to hear from anyone who's building in this space — what features matter most to you, what's the biggest time sink in your current setup, or what you'd build if the infrastructure was already handled. Happy to answer anything about the architecture or the decisions behind it.