A few weeks ago, we released WatchGrid , a lightweight Linux server monitoring service. And almost immediately, we did something that some business advisors might call counterintuitive: we open-sourced the client agent that runs on our users’ servers. You can find the repository at github.com/stheosde/watchgrid-agent .
We didn’t do this just to get a few stars on GitHub or as a cheap marketing stunt. We did it because, as developers ourselves, we have grown incredibly weary of the "black box" model of modern software. And more importantly, we believe the rise of AI-assisted coding is about to make proprietary client-side code obsolete anyway.
Here is the personal story of why we did it, and why I think more software companies should start giving away parts of their software.
Nobody wants your black box on their system#
If you build software that runs on other people’s infrastructure (whether it is an agent, a CLI tool, a library, or an integration script), you are asking for a huge leap of faith.
Think about it from the user’s perspective. Every time they install a closed-source package, run a pre-compiled binary as root, or run a setup script via curl, they are taking a security risk. They have no idea what that code is doing in the background. Is it sending environment variables back to a server? Is it leaking memory? Does it have a backdoor?
When we built WatchGrid, we knew we could not ask for that trust blindly.
If people are going to run our code on their production servers, they should be able to see every single line of it. They should be able to audit the security, compile it themselves, or even run their own forks if they do not like how we handle certain metrics.
But this is not just about security or server monitoring. This applies to almost all business and enterprise software. If you ask a client to install something inside their network or on their machines, keeping that code closed-source is a massive friction point. It slows down procurement, triggers endless security questionnaires, and breeds distrust. Open sourcing the client-facing parts of your stack solves all of this instantly.
The AI coding shift: Code is no longer a moat#
For a long time, software companies hoarded their client libraries and agents because they viewed them as intellectual property. They thought that if they open-sourced it, someone would copy the code and build a competitor.
But in the era of LLMs and AI-assisted coding, that argument is dead.
If someone wants to write an agent that checks CPU usage, Docker statuses, or queries systemd, they do not need to copy our code. They can ask Gemini or Copilot to write a highly optimized Go binary that does exactly that, and they will have it running in an hour. The code itself is no longer the moat.
The value of a software product has shifted. The moat is not the utility code that fetches data or runs on the client; it is:
- The infrastructure required to process and store that data at scale.
- The design and user experience of the central platform.
- The reliability of the alerting pipelines.
- Trust, compliance, and where the data is hosted (for us, that is Germany, which is a major point of trust for European users).
If the code itself is easy to generate, keeping it proprietary does not protect your business. It just makes it harder for your users to integrate with you.
Why open code makes your software AI-friendly#
There is another, more practical reason to open-source your code in the AI era.
Modern developers do not write integrations from scratch anymore. They point their AI coding assistants at a library or codebase and say: "Help me integrate this into my deployment pipeline," or "Write a script to automate this setup."
If your client library or agent is closed-source, AI tools are essentially blind. They cannot read the code to understand its edge cases, its command-line flags, or its internal structure. They have to rely on whatever outdated documentation they find on the web.
By open-sourcing the WatchGrid agent, we make it easy for AI tools to understand it. If you want to use an AI assistant to generate a custom Ansible playbook, write a Docker wrapper, or debug a conflict on a legacy system, the AI can read the actual source code on GitHub and give you a precise, working solution.
Open source makes your software compatible with the modern, AI-augmented developer workflow. Closed source isolates you from it.
Rethinking the "Secret Sauce"#
Open-sourcing parts of your product does not mean you have to give away the keys to the kingdom. We are not open-sourcing our entire backend platform: we still run a business, and we need to pay for our servers and our time.
But drawing a clear line between what is a client-side utility and what is platform value has been incredibly liberating. It allowed us to focus on making the backend fast, secure, and beautiful, while giving our users full control and visibility over the code running on their machines.
I hope more founders and software teams start looking at their code repositories with this perspective. Stop hoarding the utilities, the SDKs, and the agents. Open them up. It builds trust, it makes your tools easier to integrate, and in a world where AI can write the code anyway, transparency is the only real competitive advantage left.
If you want to check out our agent’s code, audit it, or tell us how to write better Go, the repo is open for business: github.com/stheosde/watchgrid-agent
And if you want to see the platform we built around it, we would love to have you try it out at watchgrid.de .
