AI Reply Generator for X
that actually reads the link.
Paste any post. Penwell opens the link it points at and reads it before writing, so the reply answers the article rather than the headline. Free, no signup, 10 a day.
Why most AI reply tools write nonsense.
A reply only earns attention if it says something the post didn't. That's the whole job. And it's the exact thing most tools can't do, because they only ever see the 280 characters in front of them. The post links to a 4,000-word essay and the tool never opens it. So it does the only thing left available: it agrees, loudly, in a slightly more enthusiastic voice than a human would use.
You've seen the output. “Such an underrated point.” “This is so true.” “Great share!” Everyone scrolling past can tell instantly that nobody read anything, because a reply that engages with the actual argument is impossible to fake — it has to contain a detail that wasn't in the post.
This page opens the link and reads the page before it writes a word. That's the only difference, and it's the entire difference.
The same post, two tools.
the best startup advice i ever got was to do things that don't scale at the start. still the most counterintuitive essay out there: paulgraham.com/ds.html
“Such an underrated point 🙌 doing things that don't scale is where all the real growth is. Great share!”
“the collison installation bit gets me every time, ‘give me your laptop’ instead of ‘we'll send you a link’ is such a small thing but it's the whole essay in one line”
The reply on the right is what this page returned for that exact post — it is the first sample button above, so you can run it yourself. The detail it leans on, the Collison brothers setting up Stripe on customers' laptops in person, sits about 1,500 characters into the essay and nowhere in the tweet. Run it again and the sentence will come out different; what won't change is that the detail is only available to a tool that opened the link.
What it reads, exactly.
This page is handed two things: the text you paste, and the first link inside it. That's the entire input. There's no account attached and no extension watching your browser, so anything it wasn't given, it doesn't have.
Links and articles
It opens the first URL in what you pasted and pulls the page's main text — blog posts, essays, docs, news — up to the first 4,000 characters. That's the opening of the argument rather than the whole piece, which is where the quotable claim usually sits; the Collison detail above is well inside it. When a link gets read, the domain is printed above the reply, so you can see it opened something.
The links it can't open
A link to another X post, because x.com serves a login shell to any server that asks — there is nothing on the other end to read, so the reply comes from your pasted text alone. A page that renders entirely in JavaScript falls back to its own summary line. And it reads one link per post, not every link in a list.
Videos
Not here. You paste text, so there is no clip to open, and a post that is a bare video with no caption is the one thing this page can't help with. Reading a clip is the connected side: the browser extension hands Penwell the video file off the post, it gets downloaded and read, and what comes back is what was said and what was on screen. No timestamps, so a reply can name what happens in the clip but not the minute it happens at. It only runs on clips whose post has almost no caption — a video with a real caption is replied to from the words — and only when X exposes the clip as a plain file. A long video served in streamed pieces has no file to fetch, and the reply falls back to the post's words and its cover frame.
Threads and X Articles
Also the connected side. There is no thread for this page to go and fetch — paste the two posts that matter and it replies to both, up to 2,000 characters. With the extension installed it is already on the page, so it passes along the last 5 posts of the conversation above the one you're replying to, plus the body of an attached X Article. Neither is something a server can fetch on its own, for the same auth-wall reason.
Frequently asked questions.
Is the AI reply generator really free?
Yes. Ten replies a day, no account, no card, no email. The limit exists so the tool stays free for everyone rather than being drained by scripts. If you need more than ten replies a day you're a serious replier, and that's what the paid product is for.
Will the reply sound like me?
Not on this page — this one writes in a neutral, plain voice on purpose. To get replies in your own voice, connect your X account and Penwell reads up to your last 150 posts first: your sentence length, your punctuation habits, the things you never do. Under about 100 posts there isn't enough there to read, so it keeps the plain voice until there is.
Do I need an X account or an API key?
Neither. Paste the text of a post and you get a reply. Nothing is connected to your account and nothing is posted anywhere — this page only writes, it never posts.
Can I use these replies commercially?
Yes. What it writes is yours. Read it before you post it, the same as anything else you'd put on your timeline.
Will replying with AI get my account banned?
This page doesn't post anything, so nothing here touches your account at all. And when you do connect an account, Penwell never posts replies through an automation API — every reply goes out from your own browser in your own logged-in session, after you approve it.