Replying on X was eating too much time.
I used to open X to “just reply to a few people.”
An hour later I was still staring at drafts.
Not because I had nothing to say — because every reply took forever. Read the post. Think of a take. Rewrite it so it doesn’t sound dumb. Soften it. Make it sound like me. Delete. Start over.
That friction adds up. Especially if you’re building in public and actually want to talk to people.
So I built XReply.
Click any post → get a reply in your voice → paste it. About 2 seconds. Not generic AI. Not “Great point!” energy. Something that sounds like you would’ve written it if you weren’t tired.
I use it myself now. Free to start (200 credits), then $15 for 5,000 — no subscription.
If replying on X feels heavier than it should, this is for you.
https://xreply.net
Would love feedback from people who reply a lot.
This addresses a real workflow problem because responding consistently can take more time than writing the original posts. The key challenge will be preserving context and personality so replies do not feel generic or automated. A strong product demo could compare an original post, a basic AI-generated reply, and the final XReply output to make the difference immediately visible. Is the product mainly designed for generating replies faster, finding the right conversations to join, or managing the entire engagement workflow?
Brilliant problem to solve! Anyone active on X knows how much time constant notifications and replies eat up. Building something that saves time while maintaining authenticity (instead of just auto-responding) is the sweet spot. This solves a real pain point for indie hackers trying to balance building + community engagement. Love it!
Nice, I was building something like this last year but for generating social media posts based on trends, though I later stopped working on it. Definitely checking out XReply, well done.
Trends-based post generation is a different beast, respect for trying it. Glad you're checking XReply out, would love to hear what you think after.
Having the 'which reply would you choose' during onboarding is a very smart move, honestly. So many tools like this have very broad onboarding flows - the one thing I've learned building hyper-personalized writer agents for content creators is taht nothing beats actually looking at examples.
The examples thing is underrated. People don't want to describe their voice, they want to show it. That onboarding insight is something I keep coming back to.
I was thinking of doing a similar project, i.e. having an To Do app that records using my voice. I will definitely use XReply and good luck to you.
This resonates — I'm early into doing this myself (posting + replying manually for a small tool I launched) and can already feel how much time it eats. Curious how XReply decides what counts as a "worth replying to" tweet — is that manual or automated?
I like that you’re solving a very specific pain point instead of trying to build another general-purpose AI writer. The focus on preserving the user’s writing style is what stands out most to me.
One suggestion would be to let new users compare an AI-generated reply with a manually written one during onboarding. That could help them quickly see how closely XReply matches their voice and build confidence in the product.
It may also help to showcase a few real before-and-after examples from actual users. Seeing authentic results often makes it easier for people to trust a tool like this.
Wishing you success with the launch!
Congrats on the launch! The "pay-per-use" credit model instead of a subscription is an excellent growth hook for indie builders—people absolutely hate subscription fatigue right now.
Your landing page copy is highly relatable. However, looking at your feature comparison table, you are missing a massive angle: Lead Generation pipelines.
For context, I recently launched LeadPulse on RapidAPI. It's a serverless API that pulls fresh local business domains and summaries in real-time. A lot of our users scrape those local business leads, find their respective X profiles, and then need to engage with them to break the ice.
If you position XReply not just as a "time-saver for browsing," but as an outbound social selling acceleration tool for agencies and B2B founders to warm up cold prospects, you can charge way more than $15 for 5,000 credits. Marketers gladly pay a premium for tools that directly speed up their sales workflows.
Quick technical question: How are you managing the fine-tuning layer to capture their "actual voice" from day one without burning massive LLM token costs?
The bottleneck almost never is generation speed, it's trust in the output. I'd track how often people paste as-is versus rewrite half of it, that tells you if this actually saves time or just produces a faster first draft to edit. Founders replying while building tend to care more about not sounding like every other AI reply than about shaving off two seconds.
You just described my old drafts folder. That whole "rewrite so it doesn't sound dumb → soften → make it sound like me → delete → start over" loop from the post was never a speed problem. It was me not trusting the draft. Paste-as-is is the only number that actually matters, and "not sounding like every other AI reply" is the exact reason I couldn't ship "Great point!" energy. Watching edit-rate closely — still don't fully trust my own read on it yet.
That funnel breakdown is useful, generated to copied to posted with edits as a branch. I'd add one more: time-to-post. A zero-edit reply can still sit unsent an hour if someone's overthinking tone. Same trust problem I deal with building FounderFlow, people don't want one more draft to edit, they want to know it's safe to just go.
“About 2 seconds” measures generation, not completion. The sharper metric is the share of replies pasted with zero edits, plus median edit time when users do change them. If the 200 free credits produce lots of generations but little paste-without-edit behavior, voice confidence is the bottleneck, not pricing.
"About 2 seconds" is the line I almost cut from the post, for this exact reason — it measures me generating, not me actually posting. Zero-edit paste rate is the real one. And you're right about the 200 credits: they're a test, not a trial. If people burn through generations and paste nothing, that's a voice problem, and no price move fixes it. That's the failure mode I'm actually scared of.
Then instrument one funnel, not a survey: generated → copied → posted, with edits as a branch. High generation/low copy means relevance failed; high copy/low post means confidence failed; high post/high edit means voice failed. That gives the 200-credit test a diagnosis, not just a conversion rate.
The interesting opportunity isn't generating replies faster—it's reducing the friction that keeps people from participating consistently on X. I'd keep validating whether users value XReply because it writes better responses or because it helps them stay present in conversations they would otherwise skip.
This is closer to why I built it than the speed pitch is. That bit in the post about building in public and "actually wanting to talk to people" — that's the real thing. Half the replies I never sent weren't hard, I just ran out of energy and scrolled past. If XReply gets you into the thread you'd have skipped, that's the win. Better wording is just the mechanism. Now I need to figure out how to measure "threads I'd have skipped" — genuinely open to ideas.
I'm glad it resonated.
The measurement question is actually the interesting part here. I have a few thoughts on how I'd think about distinguishing "better replies" from "participation that wouldn't have happened otherwise," but I don't think I'd do it justice in a thread.
I'd rather explain it in the context of XReply than reduce it to a generic engagement metric.
If you're interested, what's the best email to reach you on?
Congrats on launching. One thing I immediately wondered about is confidence in the generated replies. Saving time is a strong promise, but users will probably judge XReply by how often they can copy a suggestion and post it without rewriting it. If they still feel the need to heavily edit every response, the time savings disappear, even if the AI is technically good. Capturing a user’s voice consistently will probably matter just as much as the speed.
I specialize in independent product validation for early-stage SaaS. I evaluate products from a genuine first-time user’s perspective, uncover where trust, clarity, or usability break down, and produce professional walkthrough videos alongside structured findings that founders can use to improve onboarding, product adoption, and marketing. Want us to talk about that?
Thanks Obasekore — this is exactly the right lens.
Confidence in the reply is the product. If someone still has to heavily edit, we didn’t save them time. Voice consistency is what we’re focused on improving.
Appreciate the offer on validation too. We’re heads-down on product right now, but I’ll reach out if we want an outside pass later.