Watching your niche
See the story before it peaks
This is a radar for X, the network that used to be Twitter, and I build it on your own accounts. It searches your keywords, measures how fast each post is moving against what its author usually gets, and sends a short card to your private Telegram chat. You label every card, and after two weeks the labels show which signals came early and which were noise.
Example: made-up story
Today
Rising1 of 3 this hour
A museum's night-shift cat keeps walking into exhibit photos
- Posted
- 13:58 UTC
- First seen
- 14:02 UTC
- Alert sent
- 14:09 UTC
- What happened · AI draft
- The museum's overnight cat has turned up in three of its official exhibit photos this week.
- Why it is spreading · AI draft
- People are adding their own spot-the-cat edits, and replies are outpacing the post.
- Sources
- 3 posts, links open in XThe museum's account (example) and 2 quote posts
- Against its usual posts
- Likes 6.4x this account's usual, 11 min after posting
- Tokens already out there (optional)
- 2 found on Dexscreener, 1 paying for promotion
- Claim not checked against a second source
14:09
- Useful
- Duplicate
- Too late
- False claim
The question
this build answers
How do I get a Telegram alert when a story in my niche starts taking off on X, before it peaks?
From the first search to the label you tap.
How the radar works, step by step.
Collect
X's recent search runs on your keywords and the accounts you name, through your own X developer account. Each new post is saved with the minute the radar first saw it.
Re-check
Saved posts are looked up again by ID on a schedule. The change in likes, reposts, replies and quotes between two looks is the velocity.
Compare
Once a post clears a first filter, its author's recent posts set the baseline. That lets a small account doing six times its normal numbers stand out next to a big account's ordinary day.
Summarise
A small AI model writes two short lines: what happened and why it is spreading. Both are marked as a draft. You read them and decide, because the tool never posts, trades or publishes on its own.
Check for tokens (optional)
For token creators, Dexscreener's official search finds tokens that already exist for the story. Its orders check, which lists paid orders and boosts, shows whether one is paying for promotion. Dexscreener's own docs say paid boosts raise a token's trending score, so part of what trends is bought, and I make the card say so.
Alert and label
The card lands in your private Telegram chat with four buttons: useful, duplicate, too late and false claim. It keeps the time the story was first seen and the time the alert went out.
Running costs
What it costs to run.
These are the bills you would pay the providers directly, on your own accounts. They are not the price of the build, which I quote from your brief.
Estimates, checked 29 Sep 2026
- About $75 a month
X API, new posts found
At 500 new posts a day and $0.005 per post read. About $300 a month at 2,000 a day.
- About $100 a month
X API, authors checked
At 30 new authors a day, about $0.11 per new author checked: twenty of their posts at $0.005 each plus one profile read at $0.010.
- About $8 a month
The AI that writes the card
At 50 cards a day on Claude Haiku 4.5, which costs $1 per million tokens in and $5 out. I assume about 3,000 tokens in and 400 out per card.
- Free
Telegram Bot API
Telegram lets a bot message its users at no cost, and this one writes to a single private chat.
- About $4 to $6 a month
A small always-on server
DigitalOcean's smallest plans are $4 and $6 a month.
All together at 500 new posts, 30 new authors and 50 cards a day, on a $6 serverAbout $190 a month
Reading the same post or author again within one UTC day is normally not charged a second time; X calls this a soft guarantee. I build around it, so the bill mostly follows the new posts and new authors the radar finds. The search asks for posts only, so no author is billed until a post clears the first filter.
X's pay-per-use plan has no monthly tiers and is capped at 3 million post reads a month. The spend cap I write into the code stops a busy day from running past the budget you set. My monthly figures assume a 30-day month.
Checked at the source:
Who it is for
I designed it for people whose day starts with knowing what their corner of X is talking about. Creators and brand teams get to join a conversation while it is still small. A newsroom or a researcher can reach a story early enough to check it properly. For a community manager, a pile-on shows up as a card before it shows up in the replies.
Token creators are the fourth group, and the reason the Dexscreener step exists. They want to see a story before it peaks, and to know whether a token already exists for it and whether its rank is paid for.
What two weeks of labels tell you
Every card stores when the story was first seen, when the alert was sent and which button you pressed. After two weeks I read that back to see which signals came early, which kept arriving too late and which were noise.
Those answers set the thresholds, so the radar is tuned by your own taps rather than my guess. The records sit in a plain table you own, where a later tool, a dashboard or an analyst can read them.
The guardrails
I put a daily spend cap in the code, so a busy news day cannot run up the X bill past the limit you set. The radar reads nothing from behind a login and uses official APIs only: X's own, Telegram's Bot API and, if you want it, Dexscreener's.
Everything runs on your accounts and a small always-on server in your name. You can pause it, change the keywords or hand it to someone else without asking me.
Why take this from me
I have not built this radar for a client yet, so nobody has reviewed it. I would rather you read that here than find it out later.
What you can check is how I build. The case studies below are systems I run: scheduled jobs moving data between APIs, an assistant that cites its source instead of guessing, and one dashboard that puts the failures on top.
Before you spend anything
The fit matters.
When a saved search is enough
If your niche produces a few dozen posts a day, a saved search and X's own notifications cover it for free, and I would tell you to use those. The radar pays for its running cost when there is more than you can read and the early signal is the point.
If you want a tool that posts, replies, trades or launches a token for you, this is the wrong build. It tells you what is moving, and every decision stays with you.
Yours to keep
What lands
on your desk.
- The radar on a small server in your name, running on your own X developer account and your own Telegram bot
- Your keywords, accounts and first filter set with you, and a daily spend cap written into the code
- The Telegram card with its four label buttons, with every alert and label kept in a table you own
- The Dexscreener check switched on if you create tokens, and left out if you do not
- A written report after the first two weeks of labels on which signals came early, which were noise and what the APIs really cost
- A plain-English guide to running it, so nobody depends on me
How I build
Similar parts, already running.
Systems of mine that run every day on the same kinds of parts. Each link opens a published case study.
A few useful
answers.
What does it cost to run each month?
It depends on volume, and every figure here is an estimate I checked on 29 September 2026. X charges about $75 a month at 500 new posts a day, or $300 at 2,000, and about $100 a month to check 30 new authors a day. The AI text is about $8 a month at 50 cards a day, Telegram is free and the server is $4 to $6.
Will it post, reply or trade for me?
It does none of those. It reads, scores and sends you a card. The AI lines are marked as a draft, and every decision stays with you.
Why compare a post with its author's usual posts?
Raw numbers favour big accounts. A post doing six times its author's normal likes tells you more than a big account's ordinary day, even when the big account's totals are higher.
Do I need my own X developer account?
You do. The radar runs on your X developer account, your Telegram bot and a server in your name, so the bills and the data are yours and you can stop it at any time.
Why does the card mention paid promotion?
Dexscreener's own documentation says a paid boost raises a token's trending score. Part of what trends there is bought, so the card shows which tokens are paying before you read anything into their rank.
Make this work for you
Your tools.
Your working version.
Want this built and handed over working? Tell me what your version needs to do. One line is enough to start.
Start a brief

