Free Intensive Nov 21–22Free 2-day hands-on Vibecoding Intensive on Nov 21–22 at 16:00 MSK (Sat–Sun) · Live Cohort 3 starts Nov 23Join Free →Cohort 3 (Nov 23)
ReleaseCode / SDK · ActionableTG: @solokumi (Kumar & Solo)→ original

Kumar & Solo Moved from Paid SaaS to Their Own Server and Five Open-Source Services

The Kumar & Solo team moved off subscription services (SaaS such as Jira and Notion) and onto open-source alternatives that run on its own server. The trigger was not price but slow support, which took days to respond, and the inability to make custom changes. The team first documents the process and writes a single spec for the LLM, then runs the old and new systems in parallel for two to four weeks. You still pay for the server, but an owner and verified backups are needed from day one.

AI-processed from TG: @solokumi (Kumar & Solo); edited by Hamidun News
Kumar & Solo Moved from Paid SaaS to Their Own Server and Five Open-Source Services
Source: TG: @solokumi (Kumar & Solo). Collage: Hamidun News.
Listen to article
Article Fact Sheet & Key Figures
ParameterDetails
Primary sourceTG: @solokumi (Kumar & Solo) ↗
Editor-in-Chief
Key metric #1Docmost replaces Notion and Confluence. The team uses it to maintain procedures, instructions, and a knowledge base together, and a newcomer needs only one link.
Key metric #2Plane works in place of Jira, ClickUp, and Monday. Tasks, sprints, and the product roadmap live on your server, and you pay no separate fee for a tracker.

Many teams pay for a dozen subscription cloud services: a task tracker from one vendor, a wiki from another, and a customer chat from a third. These services are called SaaS (software as a service). When a small change is needed, support takes days to reply, and developers promise to build it "someday."

The Kumar & Solo team decided to bring these tools in-house. It set up five free open-source alternatives on a rented server of its own and uses an LLM to write whatever is missing. This approach is called vibe coding: you describe what you need, and a neural network writes the code.

This is not a scientific study but the practical experience of one team. Still, its rules will be useful to any product manager or founder tired of depending on someone else's roadmap.

How It Started

The authors first handed agents the areas where the human factor caused the most trouble: analytics, marketing decisions, call quality review, and some recurring management processes. The next step was paid B2B services. Saving money was not the original goal. The team wanted to bring these products inside its own perimeter because it was tired of waiting for support, could not customize the tools, and found that changes took a long time or never arrived. By the authors' own admission, they are difficult customers: they want a lot, and they want it immediately.

Six Rules for In-House Vibe Coding

  1. Describe the process first. Sit down with the people who actually use the tool and write down the scenarios: who does what and what they get. By the authors' estimate, about 80% of the features in old software are useless to anyone, so copying everything wholesale is not worth it.
  2. One spec file. It should contain the entities, roles, scenarios, and a separate list of what you are not doing. The LLM builds the product from this document, and you verify the result. Without a spec, you will end up with a Frankenstein within a week.
  3. Data matters more than code. Code can be rewritten in an evening, but lost data cannot be recovered. Design the database schema carefully, enable backups on the first day, and be sure to verify that you can actually restore from them.
  4. Small steps. Do one task, deploy it once, and run a check. If the model fails to fix a bug three times in a row, roll back and phrase the task differently.
  5. Run in parallel. Do not switch off the current service: let both systems run for two to four weeks and compare the results.
  6. Assign an owner. A specific person is responsible for each tool, and the repository contains a README and the spec.

Five Tools You Can Host Yourself

To get started, the authors suggest renting a server, meaning a remote computer that runs around the clock, and installing Coolify on it. Coolify works like an app store for your own server: Claude Code installs it, and after that any service from the catalog can be deployed with one click. The catalog has nearly 400 ready-made projects, including all of the ones listed below. Most of the cost is the server itself.

  • Postiz replaces SMMplanner and Buffer. It is a publishing calendar that sends posts to 30+ social networks, including Telegram and VK. You write a post once, then adjust the text and timing for each platform.
  • Chatwoot works in place of Jivo and Intercom. It brings together chat from your site, email, Telegram, WhatsApp, and Instagram in a single inbox, so a question from Instagram does not get lost among other conversations.
  • Docmost replaces Notion and Confluence. The team uses it to maintain procedures, instructions, and a knowledge base together, and a newcomer needs only one link.
  • Plane works in place of Jira, ClickUp, and Monday. Tasks, sprints, and the product roadmap live on your server, and you pay no separate fee for a tracker.

What to Consider

The approach has a cost that the original post barely mentions. You have to update, secure, and monitor your own server, and responsibility for outages falls on you rather than on a vendor. You only save money if you account for the time people spend on maintenance. That is why the rules about an owner, the README, and backup verification should be taken seriously rather than treated as a formality.

There is one more caveat: the post is partly promotional. At the end, the authors advertise their AI Superpowers course, where students rent a server together and set it up for AI agents. The five projects themselves are open source, and you can check them on GitHub.

Practical Takeaway

Start with one small tool that has few users and low cost of failure, for example a post scheduler or an internal wiki. Describe the real scenarios, write the spec, run the new service next to the old one for a month, and only then cancel the subscription. This quickly shows whether it is worth moving the rest.

Story

Key Questions Answered

How It Started?

The authors first handed agents the areas where the human factor caused the most trouble: analytics, marketing decisions, call quality review, and some recurring management processes. The next step was paid B2B services. Saving money was not the original goal. The team wanted to bring these products inside its own perimeter because it was tired of waiting for support, could not customize the tools, and found that changes took a long time or never arrived.

What are the key details on Six Rules for In-House Vibe Coding?

Describe the process first. Sit down with the people who actually use the tool and write down the scenarios: who does what and what they get. By the authors' estimate, about 80% of the features in old software are useless to anyone, so copying everything wholesale is not worth it. One spec file. It should contain the entities, roles, scenarios, and a separate list of what you are not doing. The LLM builds the product from this document, and you verify the result.

What are the key details on Five Tools You Can Host Yourself?

To get started, the authors suggest renting a server, meaning a remote computer that runs around the clock, and installing Coolify on it. Coolify works like an app store for your own server: Claude Code installs it, and after that any service from the catalog can be deployed with one click. The catalog has nearly 400 ready-made projects, including all of the ones listed below. Most of the cost is the server itself. Postiz replaces SMMplanner and Buffer.

ZK
Zhemal Khamidun — CPO at Alpina Digital, product manager of AlpinaGPT. Building AI products since 2019. Hamidun News is a pet project that grew into a production aggregator with 6,000+ readers.
✦ Unified AI Workspace · Nomad

Test Frontier LLMs & Autonomous Agents in One Workspace

Run Claude Opus 5.5, GPT-6, Gemini 3.1, and DeepSeek side-by-side in Nomad AI (nomadnet.ai) — unified model routing, deep research, and multimodal workflows under one subscription.

Try Nomad AI →Launch Nomad Agent
What do you think?
Loading comments…