OpenSERP
Self-hosted Search API for Google, Bing, Yandex, Baidu +more
I started building OpenSERP back in 2023 because I couldn't find a search API that fit my needs.
Everything I tried had the same trade-offs:
closed-source
paid per request
difficult to self-host
or unreliable once search engines started blocking requests!
So I built my own.
Over the last three years, it gradually grew from a personal scraper into an open-source project for developers building AI, SEO, and automation tools.
Today OpenSERP includes:
🌐 6 search engines (Google, Bing, Yandex, Baidu, DuckDuckGo, Ecosia)
📄 Page content extraction for LLM/RAG workflows
🔌 REST API
💻 CLI
📦 Python and JS SDKs, MCP server
☁️ Self-hosted and hosted Cloud versions
A few numbers so far:
🤞 997 GitHub stars
🍴 128 forks
🧩 SDKs for multiple languages
⭐ Public Cloud launched recently
One thing I didn't expect was how much AI agents changed the roadmap. Originally I thought SEO would be the primary audience, but today most new users are building AI agents, RAG pipelines, and automation workflows.
Building a SERP platform has also been harder than I expected:
Search engines constantly change selectors, so keeping parsers working is almost a daily job.
Browser automation is much harder than scraping HTML - you spend as much time on browser fingerprints as on the scraper itself.
Getting GitHub stars turned out to be much easier than getting paying customers.
And... people really don't open onboarding emails 😄
I'm still learning how to turn an open-source developer tool into a sustainable business, but finally opening the Cloud version to everyone feels like a big milestone.
If anyone here is building developer infrastructure or has gone through the OSS -> SaaS journey, I'd love to hear what worked (or didn't) for you.
P.S. To celebrate the launch, the public Cloud still has a 15% founder discount for early adopters.
10 Comments
nice work building this, did you build it as a solo dev? its a hefty task to put all this together and working properly, congrats getting it completed and good luck with your project. Ive recently released my first project but I find getting customers seems to be the real hurdle. Hopefully in time itll start to pick up
Thanks! 🙏🏻 It actually started as an intern project at a company back in 2020. I later polished it and released it as open source in 2023. These days I finally have partners helping me build it, especially the Cloud version.
Both comments are on the audience shift, so I'll take the part you flagged that nobody answered: 997 stars and one paying customer. That gap is the whole OSS to SaaS problem, and onboarding emails won't close it, which is why nobody opens them. Your post already contains the wedge. You said keeping parsers alive is almost a daily job and you burn as much time on browser fingerprints as on the scraper. That pain is exactly what a self-hoster feels the day a selector breaks and their agent goes dark. Cloud's pitch isn't "more engines," it's "we fight the selector and fingerprint war so your pipeline doesn't go down." Put that message where the pain happens: the repo README, the CLI output, the error a self-host throws when scraping breaks, not in an email they'll ignore. Congrats on shipping after three years, that's the hard part most people never reach.
Thanks for the advice! That's exactly the problem we're trying to solve, so users don't have to deal with selector drift, browser fingerprinting, CAPTCHAs, and all the other maintenance themselves.
And just to clarify - I didn't mention having one paying customer 🙂 The point was simply that getting GitHub stars turned out to be much easier than getting paying customers.
Fair, I misread the number, thanks for the correction. The real point still stands though: stars came easy, paying customers didn't, and that gap is the whole OSS-to-SaaS problem. A star costs someone nothing and signals "neat"; a payment signals "this saves me time or money right now". Worth digging into which of your star-gazers sit closest to that second feeling, they're a different crowd than the raw count suggests.
The shift from SEO to AI agents caught my attention.
Sometimes your earliest users aren't just customers—they're telling you what business you're actually in. I'd be cautious about treating AI agents as just another audience if they're consistently pulling the product in a different direction. That kind of pattern is often worth following before optimizing everything else.
That's exactly what happened. AI agents weren't part of the original plan, but they quickly became the primary use case.
One thing I'd keep watching is whether AI agents become your best customers or simply your first customers.
Those can look identical early on, but they often lead to very different product decisions over time.
The part that caught me was realizing AI agents ended up becoming the audience instead of the one you originally expected. Those roadmap shifts usually happen because users keep showing you something before you see it yourself.
I can easily picture a developer asking ChatGPT what people are actually using to pull search results into an agent and finding OpenSERP before they've heard the name. In that moment they're mostly looking for signs that other builders are already relying on it for real workloads.
I'm curious if you've started seeing people describe OpenSERP as "the thing behind my agent" instead of "a search API."
Not yet, but that's actually the direction I'd like to get to. Right now people mostly discover OpenSERP as a SERP API