You’ve heard the hype: AI will replace developers, automate your job, and take over your startup.
Cool. But I’m not here to sell you GPT-4 or pitch some overpriced SaaS with a .ai domain.
I run a cloud hosting company (Sliplane), and these are the five actually useful AI tools I use every single day, to ship faster, automate boring work, and avoid burning cash on things I can self-host :)
1. n8n — My Automation Sidekick
Forget Zapier. n8n is an open-source automation tool that connects APIs, databases, and even human-in-the-loop logic. I use it to run internal ops and marketing workflows.
Why I love it:
- Way cheaper than Zapier/Make (I self-host it for ~€9/mo on Sliplane)
- More flexible: you can write JS, run cron jobs, call webhooks
- Great for async workflows and internal tooling
➡️ I wrote a post about how to self-host n8n in less than 5 minutes!
➡️ I wrote a post about how I use n8n and AI agents to scale my startup if you want to steal some.
2. OpenWebUI — ChatGPT Without the Price Tag
OpenWebUI is like ChatGPT, but self-hosted. You plug in your own models (e.g. Mistral, OpenAI, etc.), and boom. A private chatbot with no fixed costs (you will still pay for API usage).
I use it for:
- Summarizing docs and meeting notes
- Brainstorming content and blog posts
- Debugging or CLI command generation
It saves me hundreds per year! And it's actually pretty good.
➡️ Here's how I set up OpenWebUI
3. Cursor — The AI-Powered VSCode I Didn't Know I Needed
Cursor is a drop-in VSCode alternative with AI superpowers. I mostly use it when I’m building CRUD things (like the API).
What it does well:
- Code suggestions that understand your context
- Inline refactors
- Can generate entire files or fix bugs
Not self-hosted, but it earns its spot by saving me a bunch of time. I still switch to Zed for more complex things where AI doesnt help that much!
If you're a student you can get it for free!
4. RunPod + a Self-Hosted Mistral Model
I use RunPod’s serverless GPU runtime to host a small Mistral model that powers parts of our backend abuse detection. Why? Because privacy matters and running it yourself is cheaper than you think!
Some cool perks:
- Pay-per-second pricing
- No vendor lock-in (OpenAI compatible API)
Also: it’s surprisingly fun to build your own AI infra.
5. OpenRouter — One API, All the Models
OpenRouter is like the Swiss army knife of LLM APIs. It lets me switch between models like Claude, Mixtral, GPT-4, and Gemini; all behind a single interface. Perfect for routing based on cost, speed, or latency.
Why I use it:
- Model fallback if one goes down
- Benchmarking different LLMs with zero code changes
- Great pricing transparency
OpenRouter is a great tool to have in your arsenal!
Wrap-Up
I don’t use AI because it’s trendy, I use it because it buys me time.
These five tools help me move faster, build smarter, and stay lean. Most are open-source. Most are dirt cheap. And all of them actually work :)
If you’re building a startup, try replacing the AI hype with something useful!
Cheers,
Jonas, Co-Founder of Sliplane
Top comments (8)
Love the emphasis on open-source and self-hosting, especially with n8n and OpenWebUI - those actually make a difference for staying lean.
Curious, have you ever hit a scaling challenge with your self-hosted stack, or has it been smooth so far?
I might be wrong, but wouldn't it be easy to create OpenWebUI with OpenRouter? I don't see openWebUI offering high availablity and performance.
Yeah you can use open router with webui!
Great tools set.. billions of token AI is energy hungry and water hogging at the same time.. youtube.com/shorts/24sPgL0CtZ8?fea...
Love the approach you took. 🙏
wow the real-world examples got me fired up to try new tools myself, you think finding your own workflow with this stuff ever gets boring
Great tips.
Does anyone know how to get api working on n8n though?
I have it linked to openwebui, its already mapped out with n8n and locally hosted ai.
I just cant actually find a tutorial for putting api into the tools the ai uses when it selects one.
If I could learn how on just one or two modules I could do any of them.
The people on the n8n discord were vague and in a private convo one of the n8n devs and some guy who seems like he knows his stuff seemed to think it was funny I couldn't figure it out. I got discouraged and stopped replying to them after the one guy tried to sell his consulting to me.
This was really discouraging for a movement that seems to encourage people of all skill sets to start vibe coding and use ai for everything.
I know a lot of stuff. I can make sites on my own without ai and all that stuff. Html, php, js, databases all the stuff. Just never worked with apis.
How is that comical? I dont know. I was really on fire for n8n for a couple of weeks and started doing other things because of this roadblock.
Any help would be appreciated if you know how to get apis to work as expected in n8n!
Some comments may only be visible to logged-in visitors. Sign in to view all comments.