DEV Community

JavaCoder7
JavaCoder7

Posted on Edited on

Selling Self-Hosted Developer Tools: How Did You Get Your First 5 Customers?

Hey developers!

I'm an indie developer building CodeGuard AI, a self-hosted AI-powered GitHub pull-request review tool built using Java 21 and Spring Boot.

Unlike SaaS products, CodeGuard AI is distributed as source code. Buyers download it and deploy it in their own infrastructure.

I've listed it for sale, but I'm still at zero sales.

I'm now trying to figure out the best way to get my first few real users.

One idea I had was to offer free evaluation licences to the first few engineering teams in exchange for honest feedback. But that brings me to a concern that I haven't figured out yet.

The source-code licensing problem

If I give someone the complete source code for a free evaluation, they can:

  • Clone and retain the entire repository.
  • Continue using it after the evaluation period.
  • Potentially redistribute or resell the code.

Even if I revoke their repository access, I cannot revoke a copy they've already downloaded.

I know piracy is a risk with downloadable software, but I don't want fear of piracy to prevent me from getting my first customers either.

I'd love to hear from people who have done this before:

  1. How did you get your first 1–5 paying customers for a self-hosted or source-code product?
  2. Did you offer free trials, discounted licences or early-access programmes?
  3. Did you share the complete source code or provide compiled binaries/Docker images?
  4. How do you handle licence expiry when customers deploy software on their own infrastructure?
  5. Do licence keys and activation systems actually work for self-hosted products?
  6. Have you ever faced piracy or unauthorised redistribution? How did you handle it?

I'm particularly interested in hearing from founders selling developer tools, self-hosted software, boilerplates, libraries or source-code products.

I'm not expecting a foolproof anti-piracy solution. I want to understand how people balance customer trust, product exposure and getting those first few sales.

Project: CodeGuard AI

Product page: CodeGuard AI on Gumroad

Would love to hear your real experiences. Even a story about how you got your very first buyer would be incredibly helpful.

Thanks!!!!!!

Top comments (5)

Collapse
 
gligorcogligorov profile image
Gligorco Gligorov •

I sell a small dev tool too, so a few honest answers from the trenches:

  1. Don't ship source code for evaluations. Ship a Docker image with a license key that expires (30 days is common). Teams that want source access can get it under a paid license later.
  2. Don't worry about piracy yet. At 0 customers, obscurity is a bigger risk than copying. The teams that would pirate it were never going to pay.
  3. Free evaluation for honest feedback is a good deal, but make it time-boxed and ask for one specific thing back (a call, or a written review).
  4. The best first buyers are people already complaining about the problem in public: GitHub issues and threads where teams ask for self-hosted or private AI code review because they can't send code to a SaaS. Replying there with something useful works much better than posting about your launch.

If it helps, I can run a search for people publicly asking for self-hosted AI code review and send you the list, free. Happy to.

Collapse
 
sweety717 profile image
JavaCoder7 •

Thanks a lot for taking the time to share such practical advice. I really appreciate it, especially the point about finding people who are already looking for self-hosted AI code review solutions.

I agree that getting CodeGuard AI in front of the right teams is much more important right now than worrying about piracy. I'm also considering a time-boxed evaluation process where teams can try the tool and share specific feedback.

And yes, if you could share that list of people or discussions looking for self-hosted or private AI code review tools, that would be absolutely awesome! It would help me reach developers who are already experiencing this problem rather than simply promoting the product to a broad audience.

Really appreciate your willingness to help. Thank you!

Collapse
 
gligorcogligorov profile image
Gligorco Gligorov •

Sorry for the slow reply, I missed your comment!

I ran the search: public GitHub issues, Hacker News, Stack Overflow and DEV from the last 60 days, plus six months of issues in the open-source AI reviewers (PR-Agent, Kodus, Mira and a few others).

The honest finding first: almost nobody asks “which self-hosted AI reviewer should I use?” in public. The demand shows up somewhere else, in teams already running a self-hosted reviewer and hitting a wall. Those are your best buyers. They’ve already decided they can’t send code to a SaaS, and the tool they picked is letting them down.

The 10 strongest:

  1. mira #69, data-residency rules, their network only allows AWS endpoints, so they need Bedrock directly: github.com/miracodeai/mira/issues/69
  2. pr-agent #2457, restricted enterprise environment, wants review to run purely locally: github.com/The-PR-Agent/pr-agent/i...
  3. kodus #2381, model gateway on a private network, and the tool refuses private addresses: github.com/kodustech/kodus-ai/issu...
  4. kodus #1126, evaluating on their own 4× H100 with vLLM, asking which model to run: github.com/kodustech/kodus-ai/disc...
  5. kodus #1161, the self-hosted model path breaks Gemini 3.x reviews: github.com/kodustech/kodus-ai/issu...
  6. kodus #2198, GitHub Enterprise Server, only a personal token with write access to every repo works: github.com/kodustech/kodus-ai/issu...
  7. kodus #1789, the self-hosted deploy template doesn't work at all: github.com/kodustech/kodus-ai/issu...
  8. kodus #2538, one step ignores their own model setting, so code goes to a provider they didn't choose: github.com/kodustech/kodus-ai/issu...
  9. pr-agent #3370, self-hosted GitLab with open-weight models behind LiteLLM, getting weak reviews: github.com/The-PR-Agent/pr-agent/i...
  10. mira #78, self-hosted Docker, can't index any repo: github.com/miracodeai/mira/issues/78

Patterns worth building for: model gateways on private networks, GitHub Enterprise Server auth, open-weight models (Qwen, DeepSeek, gpt-oss) behind LiteLLM, and an install that just works.

I have 11 more if you want them. And if it’s useful, I can keep watching these places and send you new ones as they appear, free for the next few weeks. That’s what the tool I’m building does, so you’d be helping me test it too. All I’d ask in return is honest feedback on whether the leads are any good.

Want me to set that up?

Thread Thread
 
sweety717 profile image
JavaCoder7 •

First, thank you for going this deep into the research. This is much more useful than generic advice about finding customers.

Your observation about demand showing up through issues with existing self-hosted tools is particularly interesting. It makes sense to focus on teams that have already decided to self-host but are running into limitations with model routing, private infrastructure, authentication, or deployment.

I'd definitely be interested in the additional 11 leads and having you monitor these discussions for the next few weeks.

For CodeGuard AI, I'd particularly like to understand:

  • What specific problems teams are repeatedly struggling with.
  • Whether those problems are something they would pay to solve.
  • Which gaps could be addressed by an existing solution versus requiring new features.

I'm also happy to give honest feedback on the quality and relevance of the leads for your tool.

Really appreciate you putting this together. Looking forward to exploring this further!

Thread Thread
 
gligorcogligorov profile image
Gligorco Gligorov •

Great, let’s do it!

I’ll put together a short breakdown answering your three questions (recurring problems, which ones look worth paying for, and which gaps an existing tool could close), plus the other 11 leads. Rather than posting it all in public, can you drop me a line at ggligorov1709@yahoo.com? I’ll send it there, and new ones each week from then on.