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

How it’s set up

Social-analytics tools charge a subscription to do something your own data already supports: tell you who’s worth following back. kloss_xyz did the math - $600/yr for one, $50/mo for Circleboom - and had Grok Bot replicate the useful part from his X data archive instead.

1. 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. 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. 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. 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. It keeps a backup of the original files first, so a bad call is always reversible.

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.

Why it’s cool

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.