You do not need a Fortune 500 tooling budget to ship serious AI software. In 2026, free tiers and open-source tools cover local development, backends, observability, and collaboration. Here are ten worth installing this week.
When you are ready to get paid for those skills, explore AI engineering jobs on Nova AI Jobs—including remote and LLM roles.
1. Zed (editor)
GPU-accelerated editor with snappy UX and solid collaboration features. A strong VS Code alternative when you want speed for large repos and frequent pair sessions.
2. Supabase (backend)
Postgres, auth, storage, and edge functions with a usable free tier. Excellent for AI app prototypes that need real users, not only notebooks.
3. Vercel / Netlify / Cloudflare Pages
Ship Next.js or static frontends globally. Pair with serverless functions for light AI gateways—watch cold starts and secret management.
4. Bruno or Hoppscotch (API clients)
Git-friendly or browser-fast clients for hitting model APIs and your own agent endpoints. Keep collections next to the repo.
5. Excalidraw
Default whiteboard for RAG diagrams, agent graphs, and onboarding maps. Export to PR descriptions.
6. GitHub Desktop or a visual Git client
Still valuable for complex rebases and teaching juniors. Use CLI for muscle memory; use GUI when history gets thorny.
7. Obsidian
Local Markdown notes for paper summaries, incident learnings, and interview prep. Your second brain should not be locked in a SaaS silo.
8. Sentry (developer tier)
Exception tracking for the application layer around your models. AI apps still throw normal software errors—catch them.
9. Warp (terminal)
Modern terminal UX with blocks and AI-assisted commands. Handy when juggling Docker, CUDA, and eval scripts.
10. DBeaver
Universal DB client for Postgres embeddings tables, analytics stores, and feature DBs. Community edition covers most AI side projects.
How to assemble a free AI side-project stack
- Editor + terminal for daily drive
- Supabase (or Postgres) for data + auth
- Hosted frontend for the UI
- Tracing/evals (see our evals guide) even on a budget
- Sentry for “unknown unknowns”
Ship something small, measure it, write it up. That loop beats collecting unused accounts.
FAQ
Are free tiers enough for production AI?
For early products and portfolios, yes. For scale, you will graduate—but starting free teaches cost awareness early.
What should I learn first as an AI engineer?
Solid software engineering (APIs, testing, observability) plus one applied AI path (RAG, agents, or training). Tools change; fundamentals compound.
Where can I find jobs that use this stack?
Most modern AI product teams use some mix of these. Filter by stack keywords on novaaijobs.com/jobs and study how to ramp in 90 days once you land an offer.
