Most AI dev tools just read your code and guess. Osloq actually runs it. Connect your GitHub, pick an issue, and an AI agent spins up a real sandbox, clones your repo, runs it, and tries to reproduce the bug the way a developer would. You get a report backed by real evidence. What happened, the steps it took, and whether the bug is real, not a hallucinated guess. No local setup, no "works on my machine." It handles the tedious reproduction step so you jump straight to fixing.
this came from a problem every developer knows too well. someone files a bug report, and before you can even think about fixing it, you drop what you're working on, dig through the repro steps, get your project into the exact state they described, and run it over and over just to confirm the bug is even real. half the time it ends in "works on my machine" and the issue sits there for weeks.
so i built osloq to do that part. you hand it a github issue, and it spins up a real sandbox, clones and runs your actual repo, and tries to reproduce the bug the way a developer would. then it hands you a report backed by real evidence. what it did, what happened, and whether the bug is actually real.
the hardest part while building it was trust. early on the agent would "reproduce" bugs that weren't real, basically confident hallucinations. so a lot of the work went into making it run real code and prove what it found, instead of guessing from reading the source. if it can't back a claim with evidence, it says so.
it's live today and i'd genuinely love your feedback. what would make this useful in your workflow? ask me anything, i'll be here all day.
@enessss since Osloq stops at a confirmed repro with evidence rather than proposing a fix, is the output structured enough, or is it more of a human readable triage report at this stage?
the hard part with most repro bugs I've dealt with isn't running the code, it's the missing state - a specific user's data, a race condition that only shows up under load, third party API responses that vary. how does the sandbox handle issues that need real production-like data or timing to actually trigger, or is this mostly aimed at the straightforward "steps to reproduce" kind of bug
the setup loop is the sharp edge — the agent iterates until the repo's up. but a whole class of bugs is 'doesn't build on clean checkout,' and something whose job is to get it running will quietly patch past exactly that.
Running the actual code instead of reasoning about it is the right call. Curious how Osloq handles repos with external service dependencies - database calls, third-party APIs - where the bug only shows in a specific environment setup.
Hey Enes, the reproducing part has always been the bit that quietly eats my afternoon, so seeing something take that off my plate genuinely made me pause. Feels like the kind of help I'd actually welcome.
How does the sandbox handle repos with heavy infra dependencies like docker compose or database services. Does it spin those up too or just the app code in isolation?
I would like support for retrying failed reproductions. Multiple execution attempts could separate random failures from actual bugs.
The evidence-backed part is the important line here. For a small team, a bug-repro agent is useful only if the report can become a reviewable artifact: exact assumptions, commands, env gaps, and ideally a tiny failing test or script someone can rerun. That keeps it from turning into another confident AI opinion in the triage queue.
Curious how it handles repos with external service dependencies like a database or third party API. Does the sandbox spin those up too, or do I need to provide credentials before it can run anything?
The "no evidence, no claim" stance is the right call — I build agents for commerce checkout flows, and verifying an agent's claims against what actually ran is where most of our engineering time goes too.
Two things I couldn't tell from the page: how do UI-only bugs work — if an issue only reproduces in a rendered front-end (dead button after hydration, checkout step breaking), does the sandbox drive a real browser, or is it limited to what's exercisable from CLI and tests?
And how do you keep the agent itself honest over time — do you re-run it against a benchmark of previously-verified issues when prompts or models change, so repro accuracy doesn't silently regress?
Neat concept. Does it handle bugs that need specific user state or production data to actually reproduce?
Tried it on a tricky bug I'd been putting off and the agent actually cloned the repo and walked through reproduction steps I hadn't even considered. Loved that the report felt grounded in real output rather than speculation.
the no-setup sandbox thing is such a smart move, you skip the half-day yak-shave of replicating someone else's machine just to confirm a bug. nice execution.
Reproducing the issue is genuinely the worst half of the job - half the tickets I get are 'it doesn't work' with zero context, and you burn an hour just recreating the state before you can even start fixing. Automating the boring forensic part instead of the fix itself is a smart wedge. How does it handle issues that need specific data/state to trigger? Congrats on the launch.
Really interesting approach!!!! Curious....how does Osloq handle issues that depend on external services or environment specific configurations? Anyways...congrats on the launch! 🚀
The sandbox repro thing is genuinely cool, I picked a flaky issue from one of my repos and it came back with actual logs and the failing test. Wish I'd had this last week when I burned an afternoon chasing a "bug" that turned out to be a stale local DB.
How does it handle repos that need specific env vars or a database to actually run, does it ask me to configure those or just spin up a blank sandbox and fail?
Really interesting approach. How reliable has Osloq been at reproducing tricky bugs without false positives?
About Osloq on Product Hunt
“An AI agent that reproduces GitHub issues for you”
Osloq launched on Product Hunt on July 3rd, 2026 and earned 184 upvotes and 51 comments, placing #4 on the daily leaderboard. Most AI dev tools just read your code and guess. Osloq actually runs it. Connect your GitHub, pick an issue, and an AI agent spins up a real sandbox, clones your repo, runs it, and tries to reproduce the bug the way a developer would. You get a report backed by real evidence. What happened, the steps it took, and whether the bug is real, not a hallucinated guess. No local setup, no "works on my machine." It handles the tedious reproduction step so you jump straight to fixing.
Osloq was featured in Developer Tools (515.1k followers), Artificial Intelligence (472.5k followers) and GitHub (41.3k followers) on Product Hunt. Together, these topics include over 202.3k products, making this a competitive space to launch in.
Who hunted Osloq?
Osloq was hunted by Enes. A “hunter” on Product Hunt is the community member who submits a product to the platform — uploading the images, the link, and tagging the makers behind it. Hunters typically write the first comment explaining why a product is worth attention, and their followers are notified the moment they post. Around 79% of featured launches on Product Hunt are self-hunted by their makers, but a well-known hunter still acts as a signal of quality to the rest of the community. See the full all-time top hunters leaderboard to discover who is shaping the Product Hunt ecosystem.
Want to see how Osloq stacked up against nearby launches in real time? Check out the live launch dashboard for upvote speed charts, proximity comparisons, and more analytics.
hey, product hunt! 👋
i'm enes, solo founder of osloq.
this came from a problem every developer knows too well. someone files a bug report, and before you can even think about fixing it, you drop what you're working on, dig through the repro steps, get your project into the exact state they described, and run it over and over just to confirm the bug is even real. half the time it ends in "works on my machine" and the issue sits there for weeks.
so i built osloq to do that part. you hand it a github issue, and it spins up a real sandbox, clones and runs your actual repo, and tries to reproduce the bug the way a developer would. then it hands you a report backed by real evidence. what it did, what happened, and whether the bug is actually real.
the hardest part while building it was trust. early on the agent would "reproduce" bugs that weren't real, basically confident hallucinations. so a lot of the work went into making it run real code and prove what it found, instead of guessing from reading the source. if it can't back a claim with evidence, it says so.
it's live today and i'd genuinely love your feedback. what would make this useful in your workflow? ask me anything, i'll be here all day.