bots.new_
Sign in
USE CASEMarketingModerate

“Replace a $50/mo follower analytics tool with your X data archive and a bot”

klossxyz didn't want to pay for social analytics, so he had Grok Bot replicate it from his X data archive. Hand the bot following.js and follower.js and it sorts everyone into mutuals, people you've talked to, inactives and spam - then tags who to cut. You approve every batch; it never unfollows more than 100 a day.

Grok·0.0·0 saves·9 uses·via @kloss_xyz·Tweet by @kloss_xyz
Open @kloss_xyz on X

How it runs

5 steps. Do them in order.

  1. 01**Request your X data archive** (Settings → Your account → Download an archive). It takes a while and arrives as an email link. Inside are `data/following.js` and `data/follower.js` - the two files the bot needs.
  2. 02**Hand those files to the bot** and paste the prompt below. It sorts every account into mutuals, people you've ever replied to or DM'd, and buckets like inactive, overactive/spam, not-following-back, and smart follower.
  3. 03**It tags the confident calls** - `cut now`, `cut probably`, `keep` - across both your following and follower lists, so you're triaging suggestions, not raw names.
  4. 04**You approve every batch.** Nothing is unfollowed without your say-so, capped at 50-100 a day so you never trip X's limits or nuke your graph by accident.
  5. 05**It keeps a backup** of the original files first, so a bad call is always reversible.

What you get

The insight isn't "unfollow people" - it's that the data these paid tools sell back to you is data X already hands you for free, in the archive. The bot's real job is the tedious part: parsing two big JS files, cross-referencing who you actually interact with, and turning thousands of accounts into a ranked, labeled shortlist. What makes it safe rather than reckless is entirely in the guardrails kloss wrote himself - a hard daily cap, human approval on every batch, and a backup of the originals before anything moves. That's the difference between a bot that helps you prune and a bot that quietly detonates your following list overnight. Swap in your own labels (niche, language, "engages with my posts") and the same skeleton becomes whatever social-analytics view you were about to pay for.

Prompt

Copy it. Change the names. Keep the job.

Prompt
download an archive of account data from X and wait for the email download. you'll need data/following.js and data/follower.js.

please sort and categorize all into mutuals, anyone I've ever replied or DM'd with, and use clear labels like inactive, overactive/spam, not following back, mutuals, smart follower, etc.

only I will approve the batches before any action, 50 to 100 unfollows a day. never would we go over 100 in a day.

please also help me with tagging the highest confidence suggestions with labels like cut now, cut probably, and keep for both lists.

keep a backup of all our original files too in case anything goes wrong.
Use withGrok

via @kloss_xyz · Tweet by @kloss_xyz

Tips

  • ·Request your X data archive first (Settings → download an archive; it emails you a link), then hand the bot data/following.js and data/follower.js. The prompt is kloss's verbatim - its guardrails (approve every batch, ≤100/day, keep backups) are what make it safe on a real account.
  • ·adhoc
  • ·acts-with-approval

Related setups

Same category. Different job.

More in Marketing →

Screen your X replies for burners and insults

Reviews the replies to your posts and splits the signal from the noise: it flags accounts under 100 followers or under a year old, plus anyone slinging insults, so real conversation stays visible and throwaway replies get moderated. Jason Calacanis built it to keep his mentions classy.

Marketing·Moderate·0 saves
Setup

A voice note becomes a live LinkedIn post

Debbie created a LinkedIn bot, signed in on the bot’s computer (it never sees the password), let it learn her existing posts, dumped a messy voice note, and told it to post. It posted. Receipt, not a draft-only demo.

Marketing·Moderate·0 saves
Setup