I Turned My Article idea into a micro SaaS in 7 days — Story of Convolut-AI
and fulfill my dream before my 39th birthday
This article also posted on my Medium
The Convolut app can be accessed here and on my Product Hunt launch
Introduction
I always want to create applications on the web. I've been dreaming about building web applications since my teenager. Here are some examples from my history of try to build something useful on the internet:
A trading blog on blogspot.com when I was in my early 20s
https://ticks-magnum.blogspot.com/
A small dashboard app that scrap and stream Covid19 data using Python and Tableau
But as you can see, all of these early attempts were really clumsy, and they were just either abandoned blogs or half-finished applications — using other platform infrastructure.
What I really aim for is a self-hosted, end-to-end SaaS platform that I develop with my own infrastructure and can scale.
I was however able to achieve and fulfill my dream eventually - last week, just before my 39 years old birthday! I had an idea about an application that would serve as a centralized solution to manage all of the AI contexts for users when they interact with AI.
The Problem I Was Solving
Last week, I wrote about my experience really struggling with a solution to manage my contexts because I have been working across different devices and different projects. You can find the article I wrote about this problem here:
Current Workaround Limitations
The current workaround using GitHub raw links is still okay, but there are some limitations:
Token Bloat: When the memory file grows really big, copying the whole thing will increase the number of tokens that are read by the AI.
No Filtering: There's no way to filter a particular segment or extract information from a particular project or part of the memory. You would need to go in and search for that section yourself. It’s either “Copy-Everything-Or-Nothing” kind of issue - not flexible at all.
Not for everyone: the idea of syncing your contexts memory with Github is great — only when you have programming background and have work with Git before. Try explaining this to your uncle or your friend who are not a coder and you will likely receive a raising-eye-brow looks
Output format restriction: using Github repo to store the context, I will have to store them in two versions, one Markdown / text and another version in JSON (graph-base memory) in order for me to export these file in those format, leading to duplication.
Niche Opportunity
See these challenges with the current workaround, I realized that as AI start to become a integral part of our everyday life, whether work, play, or creating something, we will need a solution that more elegant than Github.
I would predict that in 3–5 years, the number of contexts we need to work with AI assistance and tools will be increase expontentially as we “unload” the context memory to AI to free up our brain for creativity. The higher frequency we interact with AI everyday the more important this context management need become more critical. In this new AI era, context is everything.
So this would be a niche market for a SaaS tool that helps people manage their contexts online. For programmer, I can see some the example tools similar to this concept already emerged such as Context7. With a SaaS tool, we can open this capability to a lot more people to use.
(I also wrote my experience of using Base44 which you can read it here if you‘re interested in using it)
👇👇👇
And just like that — Convolut AI was born
With the help of Base44 platform and my local trusted Claude Code companion, I quickly turn my prototype into a full-scale, production-ready micro SaaS in just shy of 7 days
Here the look at its homepage
The application quickly evolved from just a Single-page-application into a sophisticated system:
Frontend and BFF (Base44-hosted)
FastAPI Backend
Self-hosted LLM micro-services
API gateways micro-services
Key Features Implemented
Supporting multi-tenants: user can easily sign-up for an account via User/Password or Google OAuth
Rich category and taggings capabilities to support multi level query and filter
Raw URL Generation: GitHub-style shareable links:
- Beautiful HTML rendering for human readability
- Allow AI chat client that have webfetch capability such as Claude, Gemini, ChatGPT etc to fetch the context via link — thus not being restricted to one device / AI platformInteract via MCP server for agent: I make a separate MCP server so AI-agent with tool-use capability like Claude, Gemma, Gemini, Qwen etc, can actually search, query, and sync context right in your chat — check out the demo here:
REST API endpoint: enhance workflow automation for power-user such as programmer who want to build custom program on their context bank or users of automated workflows platforms like PowerAutomate or n8n
Choice between using built-in LLM or Bring-Your-Own-AI: I always true believer of giving users the freedom to use their own LLM provider
I was amazed at the Speed of Modern Development
What made this rapid development possible:
— AI-Assisted Coding: Used Claude and other AI tools for rapid prototyping and debugging
— Modern Frameworks: Next.js, Supabase, and Docker eliminated months of infrastructure work. I basically standup a API server within 15 minutes
— Existing Solutions: Base44 platform provided immediate backend capabilities
— Cloud Infrastructure: VPS deployment with automated SSL certificates
The entire journey from article idea to production SaaS took just one week because I leveraged modern tools and focused on solving a real problem I was experiencing myself.
What's Next
So now I have fulfilled my dream of building a fully self-hosted micro-SaaS product. Convolut is in full production with free / paying users managing their AI contexts across devices and projects.
I'm working on additional integrations for popular productivity tools. I use the product myself and already have many ideas in mind for extension / plugins that will enhance this new context-management tool for end users productivity and creativity.
The dream of building web applications didn't take years of learning - it took identifying the right problem, using modern tools effectively, and shipping quickly.
Try Convolut and please give me feedback: Visit https://convolut.app to see how it can help you with your context management in action.